Download here - Optoelectronics Research Centre

UNIVERSITY OF SOUTHAMPTON
FACULTY OF ENGINEERING, SCIENCE AND MATHEMATICS
Optoelectronics Research Centre
Broadband Sources and Fibre Delay Line for the Functionality Enhancement of
Optical Coherence Tomography Application
by
King Nien Wong
Thesis submitted for the degree of Master of Philosophy
September 2007
UNIVERSITY OF SOUTHAMPTON
ABSTRACT
FACULTY OF ENGINEERING, SCIENCE AND MATHEMATICS
OPTOELECTRONICS RESEARCH CENTRE
Master of Philosophy
BROADBAND SOURCES AND FIBRE DELAY LINE FOR THE
FUNCTIONALITY ENHANCEMENT OF OPTICAL COHERENCE
TOMOGRAPHY APPLICATION
by King Nien Wong
The usage of supercontinuum sources has enabled a significant improvement
of tomographic imaging in optical coherence tomography (OCT) systems. However,
because of the intensity noise, the clarity of the OCT image can be sacrificed even
though much finer axial resolution can be obtained when using this technology. Using
a commercial supercontinuum SC450 (Fianium, UK) source, its performance in terms
of its interference signal shape and its relative intensity noise (RIN) was compared
with two other broadband sources, a semiconductor optical amplifier (SOA) and an
erbium-doped fibre amplifier (EDFA). As a measure of axial resolution in OCT; by
using different grating filters the full width at half maximum (FWHM) of the
interference signal and the RIN figure for each source is analysed. The measured RIN
value for the SC450 is found to be between ~26.8 dB and ~30.5 dB times noisier, in
terms of power fluctuation, when compared to the SOA and EDFA.
The construction of dispersion compensated chirped fibre Bragg gratings
(CFBGs) in an all fibre optical delay line was also discussed in this thesis. Using a
pair of identical CFBGs and positioning them in a reversed cascaded orientation
against each other, the anticipated Gaussian-like interference signal was successfully
obtained when one of the CFBGs was stretched.
The development of a time-domain OCT system was also highlighted in this
thesis. The OCT system can be used to perform a meaningful scan on small and thin
transparent samples. An OCT image of a mirror sample was presented, taken using
full bandwidth of the SOA source. Later, the SOA was amplified at higher driving
current when scanning a biological sample and the OCT image of an onion skin was
successfully constructed. Problems encountered during this research project were also
highlighted and future work to be done to try to overcome these were also outlined.
i
Contents
List of Figures
List of Tables
Declaration of Authorship
Acknowledgements
Nomenclature
Acronyms
1
Chapter 1 Overview
1.1 Broadband Sources for Optical Coherence Tomography
3
1.2 Motivation
5
1.2.1 Supercontinuum Source Characterisation
5
1.2.2 All Fibre Optical Delay Line Study
5
1.3 Arrangement of Thesis
6
1.4 References
7
10
Chapter 2 Background
2.1 Low Coherence Interferometry in Optical Coherence Tomography
10
2.2 Supercontinuum Generation
16
2.2.1 Fibre Nonlinearities
18
2.2.2 Self Phase Modulation
19
2.2.3 Stimulated Raman Scattering
20
2.2.4 Four Wave Mixing
21
2.2.5 Supercontinuum Generation via Photonic Crystal Fibre
23
2.3 Summary
24
2.4 References
25
ii
Chapter 3 Optical Coherence Tomography System Development
26
3.1 Introduction
26
3.2 Assembly and Preparation of OCT System
26
3.3 Interference Signal Detection and Multiple Samples
Characterisation
31
3.4 Doppler Frequency Observation
34
3.5 OCT Tomograms
37
3.6 Summary
41
3.7 References
42
Chapter 4 Supercontinuum Source: Performance Comparison and
43
Noise Characterisation
4.1 Introduction
43
4.2 Broadband Sources Characterisation and Comparison
44
4.2.1 Interference Signal Observation
49
4.2.2 Effects of Optical Filters Shaping to Its Coherence Function.
55
4.3 Relative Intensity Noise Measurement
57
4.4 Summary
64
4.5 References
65
Chapter 5 All Fibre Optical Delay Line for Optical Coherence
66
Tomography Application
5.1 Introduction to Optical Delay Line
66
5.2 Overview of Fibre Bragg Grating
67
5.3 Chirped Fibre Bragg Grating Characterisation
71
5.4 All Fibre Optical Delay Line Construction
73
5.5 Interference Signal Generation
75
5.6 Result and Discussion
75
5.7 Doppler Frequency Observation
77
5.8 Summary
79
5.9 References
80
iii
Chapter 6 Conclusions and Future Work
81
6.1 Summary
81
6.2 Future Work
82
6.2.1 Noise Suppression
82
6.2.2 One Micrometer Centre Wavelength Broadband Source
83
6.2.3 Interference Signal Broadening Simulation
85
6.2.4 Depth of Focus Improvement
87
6.2.5 Fibre Optic Endoscope
87
6.3 References
88
Appendix A
89
Appendix B
90
iv
List of Figures
Fig. 1.1
A typical OCT system.
2
Fig. 2.1
Basic OCT system, based on a Michelson interferometer.
10
Fig. 2.2
Relationship of OCT’s axial resolution against the FWHM bandwidth
15
of the broadband source being used.
Fig. 2.3
Broadened spectrum and its input laser pump (peak at 1064 nm). Inset
17
shows the PCF dispersion map (From [8]).
Fig. 2.4
A pulse (top curve) propagating through a nonlinear medium, e.g.
20
optical fibre, undergoes a self-frequency shift (bottom curve) due to
SPM. The front of the pulse is shifted to lower frequencies, and the
back to higher frequencies. In the centre of the pulse the frequency
shift is approximately linear (From [7]).
Fig. 2.5
Energy level diagram describing (a) Raman Stokes scattering and (b)
21
Raman anti-Stokes scattering.
Fig. 2.6
Additional frequency generated in parametric FWM process.
22
Fig. 2.7
A typical phase matching diagram where (a) a vector mismatch of Δk
22
and (b) is the perfectly phase matched case (From [10]).
Fig. 3.1
Schematic diagram of the time-domain OCT system. PC: Polarization
27
controller, PD: Photodetector, M: Mirror, OBJ: Objective lens, C:
Collimator, S: Sample and DAQ: Data acquisition card.
Fig. 3.2
OCT setup enclosed in a Perspex box for environmental temperature
28
control.
Fig. 3.3
The developed LABVIEW OCT program showing (a) the front panel
30
and (b) its block diagram.
Fig. 3.4
(a) Full spectrum of the SOA source and (b) the interference signal of
a mirror obtained using the SOA spectrum.
v
31
Fig. 3.5
Interference signal of microscope glass slide (a) scanned at 9.190 mm
32
micrometer reading (bottom layer) (b) scanned at 10.790 mm
micrometer reading (top layer).
Fig. 3.6
Interference signal of (a) plastic sheet and (b) plastic laminator.
33
Fig. 3.7
Experimental setup for Doppler frequency measurement.
34
Fig. 3.8
Doppler frequency of the OCT interference signal observed using the
36
spectrum analyser for the reference mirror driven at (a) 1 Hz (b) 2 Hz
and (c) 3 Hz, and their equivalent Fourier transform spectrum (d), (e)
and (f) using the oscilloscope, respectively. The corresponding
interference signal at their driven frequency is shown in the inset. The
Doppler frequency broadening is observed together with the increment
of the driving frequency for both the spectrum analyser’s capture and
the oscilloscope’s Fourier transform spectrum.
Fig. 3.9
An example of a cross-sectional OCT scan on a highly scattering
37
sample (reproduced from [3]). The vertical arrows denote depth
scanning. The horizontal arrows denote transverse scanning by
moving the scanning optics.
Fig. 3.10
OCT image of a mirror sample using the full spectrum of SOA source.
38
Pixel size and resolution of the image are 200 x 512 and 5 x 0.35 μm
in the transverse and axial direction, respectively.
Fig. 3.11
The power spectrum of the newly amplified SOA signal
39
Fig. 3.12
The OCT image of an onion skin using the amplified spectrum of the
40
SOA source. The images are scanned at two different positions on the
same onion skin. Both images have the pixel size and resolution of
100 x 512 and 5 x 0.24 μm in the transverse and axial direction,
respectively.
Fig. 4.1
(a) The front interface of the supercontinuum SC450 source (b)
Because the average power output from SC450 is too high, coupling
of its collimated beam into SMF has to be done by taking a partial
light beam reflected off a glass slide.
vi
45
Fig. 4.2
Full spectrum of (a) SC450 (b) SOA (c) EDFA sources and their
47
respective interference signal when launched into the OCT setup. The
3 dB bandwidths of each source are 1440.0 nm (see Appendix B for
characteristics) for the SC450, 59.1 nm for the SOA and 37.0 nm for
the EDFA. The optical spectrums of these sources are recorded with a
0.5 nm resolution bandwidth.
Fig. 4.3
Schematic diagram of signal shaping setup using chirped fibre Bragg
50
grating (CFBG) for spectral and interference signal characterisation.
Fig. 4.4
Interference signal characterization of the broadband sources using the
50
two different optically filtered spectral shapes. The OCT system is
highlighted in the shaded box.
Fig. 4.5
The filtered spectra of (a) SC450 (b) SOA and (c) EDFA sources
51
using a flat top shaped CFBG and their respective sinc shaped
function interference signals. Inset: (a) Enlarged spectral modulation
on top of the top hat spectrum. (c) Small arrow denotes the small dip
in spectrum. The filtered optical spectrums are from the OSA with a
0.01 nm resolution bandwidth.
Fig. 4.6
The filtered spectra of (a) SC450 (b) SOA and (c) EDFA sources
52
using a Gaussian shaped CFBG and their respective Gaussian shaped
function interference signals. The filtered optical spectrums are from
the OSA with a 0.01 nm resolution bandwidth.
Fig. 4.7
The SC450 spectrum that is incident on the photodetector after
53
coupling from the OCT coupler. This optical spectrum is recorded
with a 0.5 nm resolution bandwidth.
Fig. 4.8
Duality property of Fourier transform.
55
Fig. 4.9
Fourier transform of a Gaussian.
56
Fig. 4.10
Experimental setup for RIN measurement.
59
Fig. 5.1
Diffraction of UV beam upon incidence on a phase mask.
67
Fig. 5.2
Chirped grating.
69
vii
Fig. 5.3
Reflectivity spectrum and group delay for the top hat and Gaussian-
71
shaped CFBGs. The top hat grating is characterized from the short
wavelength side and has a 50 nm FWHM bandwidth with 25 ps/nm
dispersion. The Gaussian grating is characterized from the long
wavelength side and has a 23 nm FWHM bandwidth with -25 ps/nm
dispersion.
Fig. 5.4
Measurement setup for the modulation phase method.
72
Fig. 5.5
Schematic diagram of the experimental set-up. SOA: semiconductor
73
optical amplifier; PC: polarisation controller; PD: photodetector;
CFBG: chirped fibre Bragg grating. The constructed all fibre optical
delay line is highlighted in the shaded box.
Fig. 5.6
Schematic diagram of how the top hat CFBG is glued onto the fixed
74
stage and piezo stage.
Fig. 5.7
Interference signal (solid) and the Gaussian envelope fit (dash) of a
76
Fresnel backreflected signal from end-cleaved fibre in constructed all
fibre optical delay line OCT system.
Fig. 5.8
Doppler frequency of the all fibre optical delay line driven at (a) 1 Hz
78
(b) 2 Hz and (c) 3 Hz. The Doppler frequency broadening is observed
together with the increment of the driving frequency.
Fig. 6.1
OCT system with balanced photodetector.
83
Fig. 6.2
Emission spectrum of an in-house Yb ASE source.
84
Fig. 6.3
Group delay ripple of (a) top hat grating and (b) Gaussian grating.
86
Fig. 6.4
The phase ripple of (a) top hat grating and (b) Gaussian grating are
86
simulated by employing Eq. 5.8 using the group delay data in Fig. 5.3.
viii
List of Tables
Table 1.1
Examples of broadband sources used in OCT systems
3
Table 3.1
List of samples under OCT characterisation
32
Table 4.1
RIN measurement on a HP71400C Lightwave Analyser for
62
SC450, SOA and EDFA sources in the 1 MHz – 1GHz frequency
regime
Table 4.2
Measured and predicted RIN for SC450, SOA and EDFA sources
ix
63
DECLARATION OF AUTHORSHIP
I, …………King Nien Wong…………………………….declare that the thesis
entitled “..................Broadband sources and fibre delay line for functionality
enhancement of Optical Coherence Tomography”……………………and the work
presented in the thesis are both my own, and have been generated by me as the result
of my own original research. I confirm that:
•
this work was done wholly or mainly while in candidature for a research degree at
this University;
•
where any part of this thesis has previously been submitted for a degree or any
other qualification at this University or any other institution, this has been clearly
stated;
•
where I have consulted the published work of others, this is always clearly
attributed;
•
where I have quoted from the work of others, the source is always given. With the
exception of such quotations, this thesis is entirely my own work;
•
I have acknowledged all main sources of help;
•
where the thesis is based on work done by myself jointly with others, I have made
clear exactly what was done by others and what I have contributed myself;
•
none of this work has been published before submission.
Signed: ………………………………………………………………………..
Date:…………………………………………………………………………….
x
Acknowledgement
I thank my supervisor, Dr. Morten Ibsen, whose help, brilliant creativity, and enthusiasm
have guided me from the beginning until today. All the long and interesting conversations
we have had benefited me in many ways.
I thank Dr. Anatoly Grudinin, whose help, through his Fianium company, has enable me
to loan a unit of their supercontinuum source – SC450. Without it, this study would not
have been a success.
I thank my past and present research group members; i.e. Nyuk Yung Voo, Albert
Canagasabey, and Zhaowei Zhang whose input and bright discussion provided the
stimulating ideas and environment to carry forward my research.
I thank Dr. Anoma McCoy, who has helped collaboratively in the noise measurement
research. Her remarkable wide experience together with her generous assistance and
advaice has brought this research to completion with ease.
I am grateful for the help of the technicians; Simon Butler and Timothy McIntyre in their
quality workmanship crafting my project setup.
Last but not least, I would like to express gratitude to Mrs. Eveline Smith and Dr. Eleanor
Tarbox in providing smooth administrative and supervision in my study progress and
throughout my transition here.
xi
Nomenclature
c
Speed of light in the vacuum
k
Propagation constant in the vacuum
neff
Effective modal index
β
Mode propagation constant
β2
Group delay dispersion parameter
γ
Complex degree of coherence/ also nonlinear parameter
λB
Bragg wavelength
ε
Strain
λ
Wavelength
Λ
Grating pitch
ρe
Elasto-optic coefficient
τ
Time delay
ω
Angular frequency
λo
Central wavelength
Δλ
Full width at half maximum spectral width
I
Intensity
E
Electric field
P
Polarisation
Γ
Mutual coherence function
Aeff
Effective core area
α
Attenuation coefficient
vg
Group velocity dispersion
n
Refractive index
ν
Frequency
v
Velocity
fD
Doppler frequency
f
Focal length
B
xii
Δx
Focused spot size
zR
Rayleigh range
b
Confocal parameter
NAobj
Numerical aperture of the microscope objective
d
Spot size
A
Complex amplitude of the electrical pulse
P
Power
χ
Susceptibility
εo
Permittivity of free space
φ
Phase
q
Electron charge
RL
Load resistance
kB
Boltzmann constant
P
Power
B
ΔλB
Shift of Bragg wavelength
r
Mode field radii
B
xiii
Acronyms
A/D
Analog-to-digital
AC
Alternating current
APC
Angle-polished connector
ASE
Amplified spontaneous emission
CFBG
Chirped fibre Bragg grating
CMRR
Common-mode-rejection-ratio
CW
Continuous wave
DFG
Difference frequency generation
EDFA
Erbium-doped fibre amplifier
FWHM
Full width at half maximum
FWM
Four wave mixing
GDR
Group delay ripple
GVD
Group velocity dispersion
NA
Numerical aperture
NIR
Near infra-red
OCT
Optical coherence tomography
OSA
Optical spectrum analyser
PCF
Photonic crystal fibre
PSD
Power spectral density
RF
Radio frequency
RIN
Relative intensity noise
RMS
Root mean square
RSODL
Rapid scanning optical delay line
SLD
Superluminescent diode
SMF
Single mode fibre
SOA
Semiconductor optical amplifier
SPM
Self phase modulation
SRS
Stimulated Raman scattering
UV
Ultra violet
xiv
Chapter 1
Overview
Optical coherence tomography (OCT) can be used to perform in vivo, high
resolution cross-sectional imaging of microstructures in transparent as well as turbid
biological tissues [1, 2]. It is analogous to ultrasound B-scan imaging except that it
uses light instead of acoustic vibration to create the image. The term A-scan and Bscan are used commonly in OCT and ultrasonography because of the similarities in
the way both techniques are constructing their images or tomograms. An A-scan is a
one dimensional, axial profile obtained at a point location while B-scan is a two
dimensional, cross sectional image of the sample that is built up from a laterally
displaced A-scans. OCT has less penetration depth than ultrasound techniques but its
main advantage is that it can provide micron level resolution, routinely achieving
around 10 µm, which improves the ultrasound resolution more than a factor of ten [3].
OCT has great advantages over other medical imaging systems too. For example,
magnetic resonance imaging (MRI) methods are not suitable for tissue morphology
imaging and have poor resolution compared to OCT. MRI’s average resolution is only
1 mm and its imaging execution is very slow. Confocal microscopy instead does not
allow morphological tissue imaging and lacks millimetre penetration depth. Confocal
microscopy has better image resolution at 1 µm but it currently suffers from a limited
penetration depth of 200 µm.
OCT uses the correlation or interferometry techniques to measure the reflected
light intensity. Low coherence interferometry is used as it has the capability of
measuring time delay of light. Low coherence interferometry was first developed for
measuring reflections in fibre optics and optoelectronic devices [4-6]. It relies on the
interference between a split, and later re-combined, broadband optical field. The
broadband light source is coupled into an interferometer and the light is split into two
fields. The split field travels in a reference path, reflecting from a reference mirror,
Chapter 1
and also in a sample path where it is reflected from multiple layers within the sample
(Fig. 1.1). Interference between the optical fields can only be observed when the
optical path lengths of the reference and sample arm are matched within the coherence
length of the broadband source. Therefore the depth (axial) resolution of an OCT
system is determined by the temporal coherence of the light source.
Broadband
source
Fibre
coupler
Mirror
Detector
Sample
Demodulator
A/D
Converter
Computer
Fig. 1.1: A typical OCT system.
A time domain interference pattern can be obtained by translating the
reference mirror to match the optical paths of both the reference and sample arms.
A light beam is focused onto the sample and the intensity of the backscattered light
and its time delay are measured to yield an axial backscattering profile. Later a
multiple depth scan is performed by moving the scanning beam laterally, and its
backscattering profile is measured at those lateral positions to produce a twodimensional set of data.
For more comprehensive reading of OCT’s theory, latest developments and its
breakthrough applications please refer to reviews by Schmitt [7] , Fercher et al.[8]
Sampson et al. [3] and Tomlins et al. [9]. These authors have made a very thorough
review of the OCT technology ever since its inception in 1991 [1]. The applications
for this technology are extensive. To date, OCT technology has been demonstrated for
applications in biomedical fields such as ophthalmology, optical biopsy in
dermatology, dental, cardiovascular medicine, endoscopy and oncology [10, 11]. OCT
can also be adapted to other real world applications such as optical metrology
evaluation in highly scattering media. Nondestructive testing of material defects is
also prossible [12]. To date, a small number of OCT scanners have found their way to
commercialisation [13, 14].
2
Chapter 1
1.1 Broadband Sources for Optical Coherence Tomography
According to Sampson [15], one of the technologies that is having an
enormous impact on the OCT research field is the ultrahigh resolution OCT systems.
To reduce down to micron level resolution imaging, many researchers have used a
new type of broadband source, which is the supercontinuum generation based on
optical fibre technology, as it has the potential to provide a low coherence and high
power light source. Thus it is realised that the key parameter for the OCT backbone, is
Table 1.1: Examples of broadband sources used in OCT systems
Source
λo
Δλ
Emission
(nm)
(nm)
power
675
8.5
5 mW
820
20
2 mW
980
30
30 mW
1050
35
30 mW
1300
35
10 mW
1550
40
10 mW
Ti:Al2O3
776
176
20 mW
Unterhuber et al.[17]
Cr:4+:forsterite
1280
75
30 mW
Tearney et al.[18]
SLD
Reference
[16]
Kerr lens mode-locked laser
Superfluorescence fibres
Er-doped
1550 40 - 80 10–100 mW Bouma et al.[19]
Tm-doped
1800
80
7 mW
“
Nd/Yb-doped
1060
65
108 mW
Paschotta [20]
Photonic crystal fibre
725
325
27 mW
Považay et al.[21]
1100
372
100 mW
Wang et al.[22]
1000
139
n.a.
High-NA nonlinear fibre
3
Bourquin et al.[23]
Chapter 1
the selection of proper broadband sources. More on this topic will be reviewed in
Chapter 2.
Table 1.1 lists the characteristics of a variety of broadband sources suitable for
use in OCT. The most commonly used sources in OCT systems are superluminescent
diodes (SLDs). Because of their adequate level of irradiance and relatively low cost,
superluminescent sources come close to being the ideal sources for OCT imaging.
Commercially available SLDs [16] cover a wide range of wavelengths and many can
provide broad bandwidth spectra. Until recently, however, the average power and
bandwidth of these sources limited their application to slow acquisition and modest
resolution imaging.
The high power and wide bandwidth mode-locked Ti:Al2O3 laser has made it
an attractive source for fast and high resolution OCT imaging. Femtosecond laser
development has concentrated so far mainly on the temporal features of the pulses,
which were often optimized at the expense of the spectral shape. In OCT however, the
pulse duration is irrelevant, while the spectral width and shape play a crucial role.
Unlike ultrafast femtosecond time-resolved measurements, where special care must be
exercised to maintain the short pulse duration, OCT measurements depend on field
correlations rather than intensity correlations. Field correlation is preserved even if the
pulse duration is long. Moreover, its lack of portability limits its use in clinical
applications.
Much work has been done to find the best broadband light solution to produce
the best OCT image quality [24]. The best sub-micrometer ultrahigh resolution OCT
imaging ever reported to date is 0.75 μm [21]. To produce such a short coherence
length, Považay et al. uses a commercial photonic crystal fibre (PCF) pumped by a
Kerr lens, mode-locked, Ti:Al2O3 laser to generate a supercontinuum spectrum at a
mean wavelength of 725 nm and having a spectral width of 325 nm. PCF is a
promising new class of light source emitting a low temporal coherence light, similarly
to other broadband light sources. Supercontinuum generation is more efficient when
generated from PCF compared with other nonlinear media. More discussion of
supercontinuum generation will be explained in Section 2.2.
A recent review paper by Drexler [25] highlights the level of activity in
ultrahigh resolution OCT. The main activity has been the development and
deployment of supercontinuum generation sources in various forms of nonlinear
optical fibres in the OCT research area. The primary goal has been to push the
4
Chapter 1
resolution as small as possible and to reduce the complexity, cost and size of this
source. However, performance degradation is likely to be observed when a
supercontinuum broadband source is used in the OCT system as it keeps producing
intensity noise many tens of decibels above the pump noise [26]. The noise originates
from the nonlinear amplification of quantum fluctuations in the input laser light and in
the Raman scattering process within the PCF used for supercontinuum generation. In
Section 4.3, the experimental investigation will be carried out to characterize the noise
sources of a commercial supercontinuum source.
1.2 Motivation
1.2.1 Supercontinuum Source Characterisation
Recent advances in optical telecoms technology have played some part in
rolling out optical devices, suitable for OCT imaging, at a fast pace and relatively low
cost. The advent of supercontinuum sources from PCF has enabled the possibility of
acquiring ultrahigh resolution tomographic images for OCT applications in a scale of
a few microns. It is very important to know the quality outcome of OCT images using
the supercontinuum as the broadband source. This aspect of the new generation of
broadband sources has been little studied in OCT. In this thesis, early work on
characterizing the supercontinuum SC450 source (Fianium UK) performance will be
studied. Its relative intensity noise (RIN) was measured and compared against the
standard performance of two thermal-like light sources; i.e. semiconductor optical
amplifier (SOA) and erbium doped fibre amplifier (EDFA). Also, using different
grating filter shapes, their RIN value is compared. This study will be reported in
Chapter 4 of this thesis.
1.2.2 All Fibre Optical Delay Line Study
Present OCT systems have delay lines based on bulk optics. The reference arm
is controlled using various optical delay techniques such as moving mirrors, rotating
prisms, and rapid scanning optical delay (RSODL) which uses a Fourier domain
technique. To obtain a delay, the beam must come out from fibre, pass though the
bulk-optics based delay line and then recouple back to fibre. An all fibre optical delay
5
Chapter 1
line for use in the OCT system will be constructed; it will consist of two chirped fibre
Bragg gratings cascaded in opposite directions with one of the gratings being
stretched to amplify the optical path delay. The motivation of using this method is that
there is no coupling or insertion loss, there is ease of alignment, it is less bulky, more
compact and durable; all of which are inherent properties of fibre optics. This study
will be reported in Chapter 5 of this thesis.
1.3 Arrangement of Thesis
This thesis is organised in the following manner. Chapter 2 will give a brief
overview of low coherence interferometry – the key technology in OCT. Then follows
an explanation of the nonlinear effects required for the supercontinuum generation in
photonic crystal fibre. In the following chapter, the development of a time domain
OCT system in the gratings lab will be thoroughly discussed. In Chapter 4, early work
on noise characterisation in terms of interference signal performance for each SC450,
SOA and EDFA source will be reported as well as their RIN measurement using
various grating filters. In Chapter 5, the development of an all fibre optical delay line
in the OCT reference arm will be covered. Then to wrap up, in the sixth chapter, the
conclusions and the future work of this research study will be summarised.
6
Chapter 1
1.4 References
1.
D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang,
M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, "Optical
coherence tomography," Science 254, 1178-1181 (1991).
2.
J. G. Fujimoto, M. E. Brezinski, G. J. Tearney, S. A. Boppart, B. Bouma, M.
R. Hee, J. F. Southern, and E. A. Swanson, "Optical biopsy and imaging using optical
coherence tomography," Nature Medicine 1, 970-972 (1995).
3.
D. D. Sampson, and T. R. Hillman, Optical Coherence Tomography, Chapter
17 in Lasers and Current Optical Techniques in Biology (Royal Society of Chemistry,
Cambridge, UK, 2004).
4.
K. Takada, I. Yokohama, K. Chida, and J. Noda, "New measurement system
for fault location in optical waveguide devices based on an interferometric technique,"
Applied Optics 26, 1603-1606 (1987).
5.
R. C. Youngquist, S. Carr, and D. E. N. Davies, "Optical coherence-domain
reflectometry: a new optical evaluation technique," Optics Letters 12, 158-160 (1987).
6.
H. H. Gilgen, R. P. Novak, R. P. Salathe, W. Hodel, and P. Beaud,
"Submillimeter optical reflectometry," Journal of Lightwave Technology 7, 12251233 (1989).
7.
J. M. Schmitt, "Optical coherence tomography (OCT): a review," IEEE
Journal of Selected Topics in Quantum Electronics 5, 1205-1215 (1999).
8.
A. F. Fercher, W. Drexler, C. K. Hitzenberger, and T. Lasser, "Optical
coherence tomography - Principles and applications," Reports on Progress in Physics
66, 239-303 (2003).
9.
P. H. Tomlins, and R. K. Wang, "Theory, developments and applications of
optical coherence tomography," Journal of Physics D: Applied Physics 38, 2519-2535
(2005).
10.
M. E. Brezinski, Optical Coherence Tomography: Principles and Applications
(Elsevier, Burlington, MA, 2006).
11.
B. E. Bouma, and G. J. Tearney, Handbook of Optical Coherence
Tomography, (Marcel Dekker, New York, 2002).
12.
M. Bashkansky, M. D. Duncan, M. Kahn, D. Lewis, III, and J. Reintjes,
"Subsurface defect detection in ceramics by high-speed high-resolution optical
coherent tomography," Optics Letters 22, 61-63 (1997).
7
Chapter 1
13.
http://www.lightlabimaging.com/.
14.
http://www.thorlabs.com/OCT/index.cfm.
15.
D. D. Sampson, "Trends and prospects for optical coherence tomography," in
Proceedings of SPIE - The International Society for Optical Engineering(2004), pp.
25-32.
16.
http://www.superlumdiodes.com/.
17.
A. Unterhuber, B. Považay, B. Hermann, H. Sattmann, W. Drexler, V.
Yakovlev, G. Tempea, C. Schubert, E. M. Anger, P. K. Ahnelt, M. Stur, J. E. Morgan,
A. Cowey, G. Jung, T. Le, and A. Stingl, "Compact, low-cost Ti:Al2O3 laser for in
vivo ultrahigh-resolution optical coherence tomography," Optics Letters 28, 905-907
(2003).
18.
G. J. Tearney, M. E. Brezinski, B. E. Bouma, S. A. Boppart, C. Pitris, J. F.
Southern, and J. G. Fujimoto, "In vivo endoscopic optical biopsy with optical
coherence tomography," Science 276, 2037-2039 (1997).
19.
B. E. Bouma, L. E. Nelson, G. J. Tearney, D. J. Jones, M. E. Brezinski, and J.
G. Fujimoto, "Optical coherence tomographic imaging of human tissue at 1.55 μM
and 1.81 μM using ER- and TM-doped fiber sources," Journal of Biomedical Optics
3, 76-79 (1998).
20.
R. Paschotta, J. Nilsson, A. C. Tropper, and D. C. Hanna, "Efficient
superfluorescent light sources with broad bandwidth," IEEE Journal on Selected
Topics in Quantum Electronics 3, 1097-1099 (1997).
21.
B. Považay, K. Bizheva, A. Unterhuber, B. Hermann, H. Sattmann, A. F.
Fercher, W. Drexler, A. Apolonski, W. J. Wadsworth, J. C. Knight, P. S. J. Russell,
M. Vetterlein, and E. Scherzer, "Submicrometer axial resolution optical coherence
tomography," Optics Letters 27, 1800-1802 (2002).
22.
Y. Wang, Y. Zhao, J. S. Nelson, Z. Chen, and R. S. Windeler, "Ultrahigh-
resolution optical coherence tomography by broadband continuum generation from a
photonic crystal fiber," Optics Letters 28, 182-184 (2003).
23.
S. Bourquin, I. Hartl, A. D. Aguirre, P. L. Hsiung, T. H. Ko, T. A. Birks, W. J.
Wadsworth, U. Bünting, D. Kopf, and J. G. Fujimoto, "Portable broadband light
sources using a femtosecond Nd: Glass laser and nonlinear fiber for ultrahigh
resolution OCT imaging," in Proceedings of SPIE - The International Society for
Optical Engineering(2003), pp. 4-8.
8
Chapter 1
24.
A. Unterhuber, B. Považay, K. Bizheva, B. Hermann, H. Sattmann, A. Stingl,
T. Le, M. Seefeld, R. Menzel, M. Preusser, H. Budka, C. Schubert, H. Reitsamer, P.
K. Ahnelt, J. E. Morgan, A. Cowey, and W. Drexler, "Advances in broad bandwidth
light sources for ultrahigh resolution optical coherence tomography," Physics in
Medicine and Biology 49, 1235-1246 (2004).
25.
W. Drexler, "Ultrahigh-resolution optical coherence tomography," Journal of
Biomedical Optics 9, 47-74 (2004).
26.
N. R. Newbury, B. R. Washburn, K. L. Corwin, and R. S. Windeler, "Noise
amplification during supercontinuum generation in microstructure fiber," Optics
Letters 28, 944-946 (2003).
9
Chapter 2
Background
This chapter will examine the principle behind using low coherence
interferometry, utilising a broadband source in OCT and later give a brief explanation
of nonlinear effects required for supercontinuum generation.
2.1 Low Coherence Interferometry in Optical Coherence Tomography
Optical coherence tomography (OCT) is a two beam interferometric technique
which uses low coherence or partially coherent light. In a basic Michelson
interferometer case, shown in Fig. 2.1, the beam wave is split into reference field E1
and sample field E2. After reflecting back from the reference mirror and scattering
back from the sample, both fields mix on the surface of photodetector.
Reference mirror
x1
Beam splitter
Broadband
source
E1
Ein
Sample
E2
Eout
x2
Photodetector
Fig. 2.1: Basic OCT system, based on a Michelson interferometer.
Chapter 2
Optical detectors are square law intensity detection devices, where the
intensity is proportional to a time average over the electric field multiplied by its
complex conjugate. Thus the intensity that is incident on the photodetector is
∗
I out (τ ) = E out E out
(2.1)
= E1 E1∗ + E 2 E 2∗ + 2 Re{Γ12 (τ )}
{
= | E1 | 2 + | E 2 | 2 + 2 Re E1 (t + τ )E 2* (t )
(2.2)
}
= I 1 + I 2 + G12 (τ )
(2.3)
(2.4)
where I1 and I2 are mean (dc) intensities returning from the reference and sample arms
of the interferometer.
The key function in the theory of partially coherent light is the mutual
coherence function. The time average portion of the last term in Eq. 2.3 is a crosscorrelation function, which is denoted by
Γ12 (τ ) = E1 (t + τ )E2* (t )
(2.5)
and referred to as the mutual coherence function of the light field at x1 and x2.
In many cases it is convenient to work with a normalised version of the mutual
coherence function, rather than the mutual coherence function itself. The normalised
form of the mutual coherence function is defined as
γ 12 (τ ) =
Γ12 (τ )
Γ11 (0)Γ22 (0)
=
E1 (t + τ )E 2* (t )
| E1 | 2 | E 2 | 2
(2.6)
which is also known as the complex degree of coherence of the light. Take note that at
position x1, when τ = 0, self coherence of the light reduces to the intensity, Γ11 (0) =
| E1 |2 = I1. The same applies for the light field at position x2. For future reference,
γ12(0) = 1 and |γ12(τ)| ≤ 1. Eq. (2.3) can now be recast as
11
Chapter 2
I out (τ ) = I 1 + I 2 + 2 I 1 I 2 Re{γ 12 (τ )}
(2.7)
or
= I 1 + I 2 + 2 I 1 I 2 γ 12 (τ ) cos[2πυτ ]
(2.8)
which is the general interference law for the partially coherent light. The visibility, V
V=
2 I1 I 2 γ 12 (τ )
I1 + I 2
(2.9)
defines the depth or contrast of the interference fringe.
The modulus |γ12(τ)| of the complex degree of coherence is a direct measure of
the fringe visibility. In the special case, I1 = I2, the visibility and the modulus of the
complex degree of coherence are identical:
V = γ 12 (τ ) .
(2.10)
From the definition of visibility, it is clear for partially coherent light, we have
0<|γ12(τ)|<1. When γ12(τ) = 1 or 0, we have completely coherent or incoherent light
respectively.
In a two beam interferometry set-up, when two beams, I1 and I2 from the
reference and sample arm respectively, coincide at the detector, the mutual coherence
function reduces to an autocorrelation function G12(τ) (Eq. 2.4) which contains only
temporal effects. The term G12(τ) is the interferogram and it depends on the optical
time delay τ set by the position of the reference mirror. It represents the amplitude of
the interference signals that carry information about the sample under scan. The
nature of the interference signal depends on the degree to which the temporal and
spatial characteristics (coherence properties) of E1 and E2 match. Its real part
represents an interferogram with the fringes spread out
G12 (τ ) = 2 I 1 I 2 cos[ 2πυτ ] .
12
(2.11)
Chapter 2
where τ =l/c with the l as the optical path travelled by the moving mirror and c is the
speed of light.
As it is an autocorrelation function, the Wiener-Khincin theorem states that its
Fourier transform is the power spectrum S(υ) of the light source [1]. Since Γ is an
analytic signal and has no components of negative frequency:
+∞
S (υ ) = ∫ Γ12 (τ ) exp(i 2πυτ )dτ
−∞
(υ≥0)
(2.12)
thus
∞
G12 (τ ) = ∫ S (υ ) exp(−i 2πυτ )dυ
0
(2.13)
S(υ) is the transform of G12(τ), the real part of Γ12(τ) that describes the intensity
modulation, that is, the interferogram. The interferogram and power spectrum are thus
Fourier transforms of each other. From this relationship, the shape and width of the
light source spectrum are the important variable in OCT. The width of this
autocorrelation function, or the coherence length, is inversely proportional to the
width of the power spectrum.
The axial resolution is determined by the coherence length, lc of the light
source. For a source with a Gaussian spectral distribution, the axial resolution is [2]
2 ln 2 λo
Δz = l c =
π Δλ
2
(2.14)
where λo is the centre wavelength and ∆λ is the full width at half-maximum (FWHM)
spectral width. Fig. 2.2 shows the relationship of axial resolution with the choice of
different sources’ central wavelength and their FWHM bandwidth. Higher resolution
is influenced only by the source spectrum. Thus high resolution can be achieved
independently of the beam focusing condition. The mechanisms that govern the axial
and lateral image resolution in OCT are independent.
The lateral resolution in the OCT imaging system is determined by the focused
spot size, as in conventional microscopy, which is
13
Chapter 2
Δx =
4λ ⎛ f ⎞
⎜ ⎟
π ⎝d⎠
(2.15)
where d is the spot size of the objective lens and f is its focal length. In general optics,
f/d = 1/(2 NAobj) where NAobj is referring to the numerical aperture of the microscope
objective which is inversely proportional to the lateral resolution. The lateral
resolution is also related to the depth of focus or the confocal parameter, b which is
2zR, two times the Rayleigh range
b = 2zR =
πΔx 2
2λ
(2.16)
Improving the lateral resolution produces a decrease in the depth of focus, similar to
conventional microscopy.
Penetration depth of OCT is determined by the emission wavelength and
source power. In general, the OCT imaging depth of penetration is limited by both
absorption and scattering. Both of these sources of attenuation are wavelength
dependent. The presence of these two phenomena can degrade the axial resolution of
an OCT image. A common limiting factor for optical imaging of biological tissues is
the absorption profile of water in the wavelengths at 0.8 and 1.3 μm. Another setback
is the short mean scattering lengths of photons in tissue at wavelengths in the blue and
ultraviolet wavelengths. OCT imaging with a source that emits in these spectral
regions would be limited to layers of few hundred micrometres thick. However
scattering decreases monotonically in the red and near infra-red (NIR) regions as the
light experiences highly forward-directed scattering interactions in tissue. Hence
imaging in this region of the spectrum is preferable even though water absorption
increases here. Maximizing OCT imaging depth of penetration therefore requires the
use of a centre wavelength that balances these two influences of absorption and
scattering. Choosing a source with a centre wavelength of 1 μm is the most optimal, as
the influence of absorption and dispersion of water in tissue is the least at this
wavelength [3].
14
Chapter 2
30
800 nm
1100 nm
1310 nm
1550 nm
Axial resolution, Δz (μm)
25
20
15
10
5
0
0
50
100
150
200
250
300
Bandwidth, Δλ (nm)
Fig. 2.2: Relationship of OCT’s axial resolution against the FWHM bandwidth of the broadband source
being used.
As mentioned earlier, the light source is the key technological parameter for an
OCT system, and a proper choice is very important. For this purpose, the
supercontinuum source has been identified as the likely candidate for improving the
axial resolution in OCT applications. With its integral properties of very broad spectra
and high power source, the supercontinuum source is seen as having the potential for
source integration into OCT systems. In the next section, an overview of the nonlinear
effects required for supercontinuum generation will be briefly discussed.
15
Chapter 2
2.2 Supercontinuum Generation
Ever since it started being reported in 1970 [4], it has been observed that
intense picosecond pulses can exhibit spectral broadening upon nonlinear propagation
through transparent glasses and crystals. Similar broadening has also been observed in
in liquid [5] and gases [6]. The supercontinuum generation in microstructured fibre or
photonic crystal fibre (PCF) was first demonstrated by Ranka et al.[7] who reported
the generation of a supercontinuum spectrum extending from 390 nm to 1600 nm by
launching pulses of 100 fs duration, 8-kW peak power, and a centre wavelength of
790 nm into a 75 cm length of microstructure fibre with a zero dispersion wavelength
at ~ 767 nm [7].
Supercontinuum generation is a universal feature of the light-matter
interaction. Supercontinuum or white light generation is a nonlinear spectral
broadening phenomenon in the frequency domain upon launching intense ultrashort
pulses (Fig. 2.3). The origin of supercontinuum generation can be traced back to the
nonlinear Schrödinger equation. The general term for nonlinear pulse propagation in a
single-mode fibre can be expressed using the nonlinear Schrödinger equation [8]
∂2 A α
∂A 1 ∂A i
2
+ β 2 2 + A = iγ A A
+
2
∂z υ g ∂t 2
∂t
(2.17)
where z is the longitudinal coordinate along the fibre, t is the time in the reference
frame travelling with the pump light, α is the fibre loss, and A is the pulse envelope
function. The first two terms on the left-hand side describe the envelope propagation
of the pulse in fibre at the group velocity υg, while effects of the group-velocity
dispersion (GVD) are governed by β2 and the last term is the effects of fibre loss. The
nonlinear parameter γ is defined as
γ =
n2ω o 2πn2
.
=
cAeff
Aeff λ
(2.18)
The second-order nonlinear refractive index is n2 ~ 2.2 x 10-20 m2/W for silica fibres
and it is frequency independent [8]. In PCF the γ value is higher than those in
16
Chapter 2
conventional fibres. This is because of the large refractive index step between silica
and air which allows light to be concentrated in a small area, thus resulting in an
enhanced nonlinear effect. The parameter Aeff is the effective core area of the PCF
material given by
Aeff
⎛⎜ +∞ F ( x, y ) 2 dxdy ⎞⎟
∫ ∫−∞
⎠
= ⎝ +∞
4
∫ ∫ F ( x, y) dxdy
2
(2.19)
−∞
where F(x,y) is the transverse mode function. When the normalised frequency V =
2πa/λ nco2 − ncl2 is in the vicinity of 2, the effective core area is well approximated by
Aeff = πa², where a is the core radius [8]. The value of V determines the number of
confined modes that exist in the core. If the value of V for a fibre is less than 2.405, it
can be shown that only single mode propagation is satisfied.
However, Eq. 2.17 does not include the effects of stimulated Raman scattering
(SRS) and four wave mixing (FWM). The right hand side of Eq. 2.17 already models
the self-phase modulation (SPM) nonlinear optical effect.
Fig. 2.3: Broadened spectrum and its input laser pump (peak at 1064 nm). Inset shows the PCF
dispersion map. (From [9]).
17
Chapter 2
No consensus of theoretical explanations has been given for the spectral
behaviour of broadening in a supercontinuum. The processes behind supercontinuum
generation, depended particularly on pulse duration (nanosecond, picosecond or
femtosecond) and the peak power. Typically, SPM, SRS and FWM are the dominant
contributors. These processes are briefly discussed in the next few sections. It has
been found that many different operating regimes allow for supercontinuum
generation and in these various regimes the process can drastically differ.
2.2.1 Fibre Nonlinearities
When an intense electromagnetic field is applied to a material, the response of
the material depends in the nonlinear manner, upon the strength of the optical field.
The polarisation P induced by the electric dipoles can be expressed as a power series
in electric field E as
(
)
P = ε 0 χ (1) ⋅ E + χ ( 2 ) : EE + χ ( 3) M EEE + ... ,
(2.20)
where ε0 is the permittivity of free space, and χ(j) (j = 1,2,…) is the jth order
susceptibility and a tensor rank j + 1 of the medium. For silica fibre, an isotropic
centrosymmetric material, the number of independent terms in the third order
susceptibility χ(3) is reduced to only three [8]. For centrosymmetric molecules only
odd-order susceptibilities are non-zero. The third order susceptibility χ(3) can be nonzero for centrosymmetric and noncentrosymmetric materials. As a result, the lowest
order nonlinear phenomena in silica fibre originate from χ(3). The second-order
susceptibility χ(2) however vanishes for centrosymmetric material. χ(2) can gives rise to
nonlinear effects such as sum and difference frequency generation and second
harmonic generation. The χ(3) term is responsible for nonlinear effects such as third
harmonic generation, two photon absorption, four wave mixing and nonlinear
refractive index [8]. The FWM and SRS nonlinear effects covered in this section
originate from this third-order susceptibility χ(3).
18
Chapter 2
2.2.2 Self Phase Modulation
When a high optical intensity light is launched into an optical fibre, it will
cause a nonlinear phase delay which has the same temporal shape as the optical
intensity. This effect produces a refractive index change with intensity, I
n( I ) = no + n 2 I
(2.21)
where no is the linear refractive index, and n2 is the second-order nonlinear refractive
index of fibre. When the optical pulse is travelling in the fibre, this will cause a timedependent phase shift according to the time-dependent pulse intensity. For this, the
initial unchirped optical pulse acquires a chirped shape; a temporally varying
instantaneous frequency. For a Gaussian pulse, propagating the distance of L, the
instantaneous frequency ω(t) is given by [8]
ω (t ) = ω o +
⎛ t2
4πLn 2 I
⎜⎜ − 2
t
exp
⋅
⋅
λT 2
⎝ T
⎞
⎟⎟
⎠
(2.22)
where ωo is the carrier frequency of the pulse and T is the half-width (at 1/e intensity
point). In practice it is customary to use the FWHM in place of T. Plotting ω(t) shows
the frequency of each part of the pulse (Fig. 2.4).
In the normal dispersion regime, lower frequencies or ‘redder’ portions of the
pulse travel faster than higher frequencies or ‘blue’ portions. Thus, the front pulse
moves faster than the back, which broadens the pulse in time. In the regions of
anomalous dispersion, the pulse is compressed temporally and becomes shorter. This
effect can be exploited to produce ultrashort pulse compression and soliton effects.
19
Chapter 2
Fig. 2.4: A pulse (top curve) propagating through a nonlinear medium, e.g. an optical fibre, undergoes a
self-frequency shift (bottom curve) due to SPM. The front of the pulse is shifted to lower frequencies,
and the back to higher frequencies. In the centre of the pulse the frequency shift is approximately linear
(From [8]).
2.2.3 Stimulated Raman Scattering
Stimulated Raman Scattering (SRS), a third-order susceptibility χ(3) is an important
nonlinear phenomenon that can lead to the generation of new spectral lines. By
introducing the Raman effect for pulse evolution in SMF, SRS process can be
included in Eq. 2.17 by modifying it to the generalised (or extended) nonlinear
Schrödinger equation [8]. SRS results from a stimulated inelastic scattering process in
which the optical field transfers part of its energy to the medium and generates a
photon.
The spontaneous Raman effect can be observed when a beam of light
illuminates any molecular medium and the scattered light is observed
spectroscopically. The Raman effect scatters only a small fraction of the incident
optical field into other fields. Those new frequency components are shifted to lower
frequencies and are called the Stokes line and those shifted to higher frequencies are
called the anti-Stokes lines. The amount of frequency shift is determined by the
vibrational modes of molecules.
20
Chapter 2
These properties of Raman scattering can be understood quantum
mechanically through the use of energy level diagrams shown in Fig. 2.5.
n’
n’
ωp
ωS
ωp
n
g
ωa
n
g
(a)
(b)
Fig. 2.5: Energy level diagram describing (a) Raman Stokes scattering and (b) Raman anti-Stokes
scattering.
Raman Stokes scattering can be described as an incoming photon of frequency ωp that
encounters one of the molecules in ground state g and excites the molecule to make a
transition from the ground state to a virtual intermediate level associated with the
excited state n’, followed by a transition from the virtual level to the final lower
excited state n as the molecules emits a lower-frequency photon ωS. Anti-Stokes
Raman scattering consists of a transition from level n to level g, with n’ serving as the
intermediate level, resulting a higher-frequency photon ωa. The anti-Stokes lines are
typically orders of magnitude weaker than the Stokes lines. For PCF, when the pump
power exceed a threshold value, a stimulated version of Raman scattering can occur in
which the Stokes line grows rapidly inside the medium with high conversion
efficiency. The lower energy Stokes lines are generated on the red-side of the
spectrum whereas the anti-Stokes lines are shifted to the blue-side of the spectrum. In
practice, the Stokes wave can serve as a pump to generate a second-order Stokes wave
if its power exceeds the threshold value.
2.2.4 Four Wave Mixing
Four wave mixing (FWM) is a parametric process in the optical fibre which
can be viewed as two photons of energies ћω1 and ћω2 that are annihilated and two
new photons of energies ћω3 and ћω4 that are created such that the net energy is
conserved during the process. It is a process where no energy is transferred to a
21
Chapter 2
medium. Each of the four waves has its own direction of propagation, polarisation and
frequency. For this process to occur, it requires the phase matching condition to be
satisfied for momentum conservation. The parametric interactions are strongest when
waves are phase-matched. Another equivalent way to understand this is by realizing
that, since the energy transfer is a coherent process, all four photons or waves must
maintain a constant phase relative to the others in order to avoid any destructive
interference.
ωa
ωS
ωp
Frequency
Fig. 2.6: Additional frequency generated in parametric FWM process.
Consider this process where two pump photons of equal frequency create a Stokes and
anti-Stokes photon (Fig. 2.6) such that
ω p + ω p = ω S + ωa .
(2.23)
In a single mode fibre, when all four photons propagate in the fundamental mode in
the same direction, the phase matching condition requirement can be written as
Δk = k1 + k 2 − k 3 + k 4 .
(2.24)
where Δk is the phase mismatch. Fig. 2.7 shows the pictorial representation of the
wave vector mismatch: situation (a) shows a finite wave vector mismatch, while (b)
demonstrates the corresponding phase matched case.
k4
k2
k2
Δk
k1
k3
k4
k1
(a)
k3
(b)
Fig. 2.7: A typical phase matching diagram where (a) a vector mismatch of Δk and (b) is the perfectly
phase matched case (From [10]).
22
Chapter 2
2.2.5 Nonlinear Supercontinuum Generation in Photonic Crystal Fibre
Supercontinuum generation is more efficient in PCF compared to other
nonlinear media. PCFs, in this example, are made up of silica where the core is
surrounded by an array of microscopic air holes stretching along its entire length. The
large refractive index step between the silica core and its air hole cladding enables
light to be concentrated onto a very small area, hence resulting in enhanced nonlinear
effects. Tight focusing of intense pump light thus increases the peak power and leads
to broadening of the input spectrum or supercontinuum generation. For input pulses
that are not too short (pulse width >1ns) and not too intense (peak power <10mW),
the fibre plays a passive role (except for pulse energy losses) and acts as a transporter
of optical pulses from one place to another [8]. In this case, its shape or spectrum will
not be significantly affected. However as pulses become shorter or more intense,
different effects of nonlinearity regime kick in.
SPM can lead to the increasing of the bandwidth due to nonlinear refractive
index. When the pump wavelength of femtosecond pulses is close to the zero
dispersion wavelength of the PCF, the low-power spectral broadening seems to be
dominated by SPM. However, it has been demonstrated that femtosecond pulses are
not necessarily needed for efficient supercontinuum generation in a PCF [11]. Using
nanosecond and picosecond excitation instead, the role of SPM is limited because
these pump pulses are long and the power sufficiently low to prevent the self-focusing
effect. SRS will firstly dominate, in this case, and it will extend the spectrum at the
long wavelength side of the pump, generating a broadband of frequencies between the
pump wavelength and the zero dispersion wavelength [12]. In the second stage, these
SRS generated frequencies serve as parametric pumps to continuously amplify the
generated blue-shifted light of shorter wavelength through FWM. This mechanism is
effective because phase matching is achieved regardless of the pump power and SPM
has a negligible influence. In FWM, two photons are summed to produce two new
photons of different frequencies. Efficient FWM requires that the process be phasematched; i.e. Δk = 0.
23
Chapter 2
2.3 Summary
In this chapter, it has been reviewed how the mechanism of OCT system
works using the low coherence interferometry technique. Thus in the following
Chapter 3, the development and demonstration of a time-domain OCT system will be
reported, where its construction will be based on the basics reviewed earlier in this
chapter. It has been showed too how supercontinuum spectrum from PCFs can be
generated using the nonlinear effects of SPM, SRS and FWM. Its broadened white
light will be used extensively in noise characterisation experiments, as discussed in
Chapter 4.
24
Chapter 2
2.4 References
1.
W. H. Steel, Interferometry (Cambridge University Press, 1983).
2.
D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W. Chang,
M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G. Fujimoto, "Optical
coherence tomography," Science 254, 1178-1181 (1991).
3.
Y. Wang, J. S. Nelson, Z. Chen, B. J. Reiser, R. S. Chuck, and R. S. Windeler,
"Optimal wavelength for ultrahigh-resolution optical coherence tomography," Optics
Express 11, 1411-1417 (2003).
4.
R. R. Alfano, and S. L. Shapiro, "Observation of Self-Phase Modulation and
Small-Scale Filaments in Crystals and Glasses," Physical Review Letters 24, 592
(1970).
5.
W. L. Smith, P. Liu, and N. Bloembergen, "Superbroadening in H2O and D2O
by self-focused picosecond pulses from a YAlG: Nd laser," Physical Review A 15,
2396 (1977).
6.
P. B. Corkum, C. Rolland, and T. Srinivasan-Rao, "Supercontinuum
Generation in Gases," Physical Review Letters 57, 2268 (1986).
7.
J. K. Ranka, R. S. Windeler, and A. J. Stentz, "Visible continuum generation
in air-silica microstructure optical fibers with anomalous dispersion at 800 nm,"
Optics Letters 25, 25-27 (2000).
8.
G. P. Agrawal, Nonlinear Fiber Optics (Academic Press, San Diego,
California, 2001).
9.
M. Rusu, A. B. Grudinin, and O. G. Okhotnikov, "Slicing the supercontinuum
radiation generated in photonic crystal fiber using an all-fiber chirped-pulse
amplification system," Optics Express 13, 6390-6400 (2005).
10.
J. F. Reintjes, Nonlinear Optical Parametric Processes in Liquids and Gasses
(Acedemic Press, Orlando, 1984).
11.
S. Coen, A. H. L. Chau, R. Leonhardt, J. D. Harvey, J. C. Knight, W. J.
Wadsworth, and P. S. J. Russell, "Supercontinuum generation by stimulated Raman
scattering and parametric four-wave mixing in photonic crystal fibers," Journal of the
Optical Society of America B: Optical Physics 19, 753-764 (2002).
12.
S. Coen, A. H. L. Chau, R. Leonhardt, J. D. Harvey, J. C. Knight, W. J.
Wadsworth, and P. S. J. Russell, "White-light supercontinuum generation with 60-ps
pump pulses in a photonic crystal fiber," Optics Letters 26, 1356-1358 (2001).
25
Chapter 3
Optical Coherence Tomography
System Development
3.1 Introduction
This chapter will describe the development of a time-domain, fibre-based OCT
system. Before the OCT system is able to generate OCT images, there are two major
problems that need to be solved. First, the optical path lengths of the reference and
sample arm need to be matched. For this, the length of the fibre used in the arms will
be carefully measured and cut. Secondly, the sampling data acquired by the computer
need to be synchronised, processed and fine controlled with the scanning optics
movement so that a full complete image or tomogram can be constructed. For this,
software will be developed to handle all these requirements.
3.2 Assembly and Preparation of OCT System
A simple time-domain OCT setup was designed and built, the schematic of
which is shown in Fig. 3.1. A picture of the setup is shown in Fig. 3.2. The fibre based
OCT system was chosen for a number of reasons [1]; compactness and durability,
ease of beam coupling and alignment and its relatively wide acceptance in OCT
research. Fibre optic implementations of OCT can also be used in conjunction with
endoscopes or catheters to provide physicians with imaging of sub-surface tissue
morphology.
Chapter 3
PC
Broadband
source
C
M
OBJ
S
Reference
arm
3dB coupler
C
PD
Sample
arm
Pre-amplifier
Band-pass
filter
Rectifier
Low pass
filter
A/D
Converter
(DAQ)
OCT Image
(LABVIEW)
Fig. 3.1: Schematic diagram of the time-domain OCT system. PC: Polarization controller, PD:
Photodetector, M: Mirror, OBJ: Objective lens, C: Collimator, S: Sample and DAQ: Data acquisition
card.
Fig. 3.1 shows the schematic of a simplified fibre-based OCT system, which is
divided into two parts, the optical part (upper box) and the electrical part (lower box).
In the optical part, light from the broadband source is launched and split in half at the
fibre coupler. The split light, reflected back from the reference arm and the sample
arm, then later interfere at the coupler and are detected at the photodetector. The
probe at the sample arm collects the detailed information on the sample. The electrical
part of the OCT system translates the information into either a 2-dimensional or a 3dimensional image.
The optical transmittance, coating, and wavelength-dependent losses of the
fibre optics employed and the delivery system optics can strongly influence the axial
resolution as well as the sensitivity of the OCT system.
27
Chapter 3
Fig. 3.2: OCT setup enclosed in a Perspex box for environmental temperature control.
Single-mode fibres (SMFs) with appropriate cut-off wavelengths have to be
used to provide single-mode light propagation. For this setup, commercial optical
fibre (SMF-28, Corning) was used. Conventional fibre couplers are designed to
maintain 3 dB splitting over a wavelength range of typically ±10 nm. When using
these fibre couplers for delivery of broad-bandwidth laser light, unequal beam
splitting with respect to wavelength and power can occur and will reduce the
resolution. Hence a broad bandwidth, wavelength flattened, 3 dB fibre optic coupler
operating at 1550 nm wavelength was used. This is to maintain the broad bandwidths
and consequently better axial resolution. It is important to note that bidirectionality of
fibre-based beamsplitters is important in order not to introduce wavelength-dependent
losses on the way back to the detector.
The low coherence interferometry is based on the occurrence of fringes if the
optical path lengths of the reference and sample arms coincide within the size of
coherence length. The length of the fibre arms are carefully measured and cut to
match its coherence length. Angle polished connector (APC) patch cords are used in
both arms to minimise backreflections. The polarisation controller is inserted in the
reference arm to compensate for the polarisation mismatch in the two arms. The
28
Chapter 3
polarisation of light can be caused by sample inhomogeneity, environmental
temperature changes, connectorisation and bending of both fibre arms. Introducing the
polarisation controller will allow the manual adjustment of the polarisation in the
reference arm, and thus improve the sensitivity of the OCT system.
Light in the reference arm is reflected from an Al-coated mirror (reflectivity
>90% at 1550 nm) which is mounted on a piezo stage (17AMP001, Melles Griot).
The piezo stage was driven with a triangular waveform to simulate a linear scanning
mechanism. The piezo stage is able to perform ~175 μm forward and backward
translation and the velocity of the stage was set to 350 μm/s at 1 Hz driving frequency.
This scanning range and velocity is relatively short and slow, respectively, compared
with that of a state of the art OCT system. A rapid scanning optical delay line
(RSODL) which uses Fourier domain technique can enable a speedier scanning in the
range of several kHz and video rate imaging [2]. But then again, this and the earlier
mirror techniques use bulky and expensive mirror optics. To compensate for these
disadvantages, a new all fibre optical delay line is proposed and this will be covered
in Chapter 5.
In the sample arm, the installed collimator (f = 11 mm, NA = 0.25) is able to
wholly collect the light from the APC path cord as its numerical aperture is higher
than the numerical aperture of the SMF, of 0.14. This beam is not entirely collimated,
as the numerical aperture of the APC partch cord and the aspheric lens on the
collimator is not the same. However, since the distance separation of the collimator
and the objective lens is very near, the collected beam will be assumed collimated.
The collimated beam is focused onto the sample using an objective lens (f = 23.5 mm,
NA = 0.22) with low numerical aperture value. The beam is able to focus down onto a
spot size, Δx = 4.3 μm (Eq. 2.15). This beam size, incident on the sample, determines
the transverse or lateral resolution of the OCT image. The depth of focus, b, on the
other hand is limited to 18.59 μm (Eq. 2.16). The common OCT technique is able to
produce millimetre depth of penetration (approximately 2-3 millimetres in tissue).
Thus, there is still plenty of room to scale up the current OCT system. The methods to
improve this depth of focus will be discussed in Chapter 6. The mechanical
performance of the scanning optics used for this transverse and depth scanning has to
be accurately selected and correctly controlled. Mechanical jitter or displacement of
adjacent depth scans as well as noisy control signals of the scanner might result in
distorted, and therefore resolution degraded, OCT tomograms. A high performance
29
Chapter 3
translation actuator (850G-HS, Newport) was used, which is able to shift the scanning
optics at the resolution of 1 μm.
In the detection part, the photodetector (D400FC, Thorlabs) collects the
interference signal, which contains information of the sample along with undesired
noise. The received signal is weak and hence a pre-amplifier is used to boost up the
signal power. A band-pass filter is used to filter out the undesired noise with the bandpass centre frequency locked at the Doppler frequency. A/D converter (PCI-MIO16E-4, National Instrument) or a DAQ card is used to convert the analog signal into a
12-bit digital signal, which will be interfaced into a computer. The signal will then be
demodulated to obtain the envelope of the fringe. The process of demodulation is
currently being done offline using an in-house written MATLAB program. The
demodulation process involves full-wave rectification of the signal, followed by low
pass filtering and then storing of its final envelope function. The piezo stage motion
control (17PCZ013, Melles Griot) in the reference arm and the 2-axis scanning optics
actuator control (MM4005, Newport) in the same sample arm, together with the
digitized signal were controlled by the LABVIEW program, as shown in Fig 3.3. The
flow chart of the LABVIEW program is shown in the Appendix A.
This OCT system will serve as a foundation for experiments reported
hereafter.
(a)
(b)
Fig. 3.3: The developed LABVIEW OCT program showing (a) the front panel and (b) its block
diagram
30
Chapter 3
3.3 Interference Signal Detection and Multiple Samples Characterisation
Using a low coherence ASE signal, generated from an SOA (CQF874-308C,
JDS Uniphase), operating at a centre wavelength λc = 1535.13 nm and having a 3 dB
spectral bandwidth of 59.09 nm, its interference signal is obtained and detected (Fig.
3.4). The source signal is launched at full power of + 0.41 dBm. Theoretically, the
axial resolution obtained using Eq. 2.14 would give a FWHM length of approximately
17.5 µm. Using a mirror as a sample under test, the FWHM of its interference signal
is measured to be 17.6 ± 0.1 µm, which conforms to the coherence length of the SOA
source.
1.0
(a)
FWHM
~17.6 ± 0.1μm
Intensity (A. U.)
Intensity (A.U.)
0.8
0.6
(b)
1.0
FWHM
~59.09 nm
0.4
0.2
0.5
0.0
-0.5
-1.0
0.0
1450
1500
1550
1600
1650
-20
Wavelength (nm)
-10
0
10
20
Delay (μm)
Fig. 3.4: (a) Full spectrum of the SOA source and (b) the interference signal of a mirror obtained using
the SOA spectrum
To demonstrate the functionality of this OCT system to scan, three different
transparent samples were used for a preliminary test. In the future, those samples can
be used for system calibration before a meaningful OCT image scan can be performed
on other inhomogeneous samples such as a biological tissue. The SOA was again used
as the light source. The samples are listed in Table 3.1 together with their known
thickness, which is measured using a micrometer. The samples are then placed under
the scanning optics before each scanning is performed.
31
Chapter 3
Table 3.1: List of samples under OCT characterisation
Sample
0.4
Refractive index
Microscope glass slide
1.45
1.050
Plastic sheet
1.46
0.055
Plastic laminator
1.46
0.105
0.4
(a)
(b)
0.2
Voltage (V)
0.2
Voltage (V)
Thickness (± 0.005 mm)
0.0
0.0
-0.2
-0.2
-0.4
-0.4
-20
0
20
-20
40
0
20
40
Delay (μm)
Delay (μm)
Fig. 3.5: Interference signal of microscope glass slide (a) scanned at 9.190 mm micrometer reading
(bottom layer) (b) scanned at 10.790 mm micrometer reading (top layer)
As the thickness of the microscope glass slide is beyond the scanning range of
each reference mirror sweep (± 175 µm), its interference signal is captured at two
different focusing distances at the scanning probe. The scanning probe is affixed with
a micrometer adjuster and can be manually adjusted. Its interference signals are
shown in Fig. 3.5. From the two micrometer readings, the glass slide thickness is
calculated to be (10.790 - 9.190) mm/1.45 = 1.100 mm. The calculated value does not
differ much from the glass slide’s real thickness (1.050 mm). The bottom layer
produced a weaker backreflected fringe, as compared to the top layer, because the
light signal will attenuate substantially while travelling to and fro the bottom layer.
The plastic sheet and plastic laminator used are not uniformly flat on their
surfaces. From the interference signal in Fig. 3.6, two fringes can be observed. The
bigger fringe of the two is the stronger backreflected light from the top layer of the
plastic. From the two fringe distances in Fig. 3.6(a), the plastic sheet thickness is
calculated to be 79.5 µm/1.46 = 54.5 µm. For the plastic laminator sample, from its
32
Chapter 3
two fringe separation distance, its thickness is thus calculated to be 150.0 µm/1.46 =
102.7 µm. These values again do not differ much from their exact measured thickness
value. From all these observation, the constructed OCT system is shown to be able to
perform a meaningful one dimensional axis scan.
1.0
1.0
(a)
(b)
~150.0 ± 0.1µm
~79.5 ± 0.1µm
0.5
Voltage (V)
Voltage (V)
0.5
0.0
0.0
-0.5
-0.5
-1.0
-1.0
-60
-40
-20
0
20
40
-80 -60 -40 -20
60
0
20
Delay (μm)
Delay (μm)
Fig. 3.6: Interference signal of (a) plastic sheet and (b) plastic laminator
33
40
60
80
Chapter 3
3.4 Doppler Frequency Observation
The use of a reference mirror, as a delay line in OCT, is common in order to
shift the coherence gate along the sample depth to achieve an interference signal. In a
time-domain OCT system, the delay line will introduce a time delay, τ, at a velocity,
v, and simultaneously generate a corresponding heterodyne frequency. The constant
velocity movement of the mirror will shift the centre frequency of the interference
signal to the Doppler frequency
fD =
2v
(3.1)
λc
where λc is the source centre wavelength used. Take, for example, a mirror secured on
a piezo driven stage forward scanned at 175 μm and oscillating at 1 Hz frequency,
using an SOA source with FWHM spectral width of 59.10 nm, where its centre
wavelength is at 1535.13 nm, the Doppler frequency, fD, is calculated to be 455 Hz.
Thus for mirror scanning at 2 Hz and 3 Hz, it will produce a Doppler frequency at 909
Hz and 1364 Hz respectively.
PC
SOA
C
M
OBJ
M
Reference
arm
3dB coupler
Marconi RF
spectrum
analyser or
oscilloscope
C
Photodetector
Fig. 3.7: Experimental setup for Doppler frequency measurement
Three Doppler frequency measurements were taken. Using the time-domain
OCT setup in Fig. 3.7, the Doppler frequency was measured using an RF spectrum
analyser (Model 2382, Marconi) and by taking a Fourier transform of the interference
34
Chapter 3
signal on a oscilloscope(TDS3012B, Tektronix). The interference signal is detected
using a photodetector (Model 1811, New Focus) before feeding the electrical signal
into the spectrum analyser or the oscilloscope. The video bandwidth and resolution
bandwidth for the spectrum analyser was maintained at 10 Hz and 30 Hz respectively.
The signal attenuation into the spectrum analyser is 40 dB. Power incident on the
detector was maintained at -14.80 dBm in order not to saturate it. For the piezo stage,
a triangular waveform was given at three different frequencies, i.e.: 1 Hz, 2 Hz and 3
Hz. The captured RF signal from the spectrum analyser and its equivalent Fourier
transform spectrum from the oscilloscope is shown in Fig. 3.8. All the three
measurements on the spectrum analyser are subject to 16 times sample sweeping to
smooth out the measurement and to reduce noise before being recorded.
The peak frequencies in Fig. 3.8 (a), (b) and (c) were observed at 480 Hz,
1000 Hz and 1570 Hz for the mirror scanning at 1 Hz, 2 Hz and 3 Hz respectively.
The difference in the frequency value with the actual Doppler frequency is due to the
turning point at which the stage is moving in the opposite direction. Even though the
peaks of the frequency are still centred at the Doppler frequency as the mirror
scanning rate is increased, frequency broadening is observed together with this
increment. Besides that, distinct peaks appeared alongside the broaden frequency
range in the spectrum analyser. Besides the Doppler frequency, the interference signal
also generated other frequency components alongside it. This means that by
increasing the scanning rate, the interference signal can get noisier. These evident
peaks may somehow have appeared from the resonance frequency of the mirror
driven by the piezo stage. In Fig. 3.8 (b), each of these peaks is a distance apart at ~50
Hz. The alternating current (AC) power line may have contributed to the peaks’
appearance too. An unstable centre frequency generated from the function generator
that is used to drive the piezo stage may have contributed to this frequency
broadening.
For the Fourier transform spectrum, the peak frequencies in Fig. 3.8 (c), (d)
and (e) were observed at 455 Hz, 935 Hz and 1440 Hz for the mirror scanning at 1
Hz, 2 Hz and 3 Hz respectively. These values do not differ much from the theoretical
Doppler frequency calculated earlier. In this frequency domain, multiple harmonics
are recorded along side the main Doppler frequency. This is because the number of
points when the Fourier transform is performed is under sampled, an effect of
aliasing.
35
0
(a)
(d)
-20
-40
Electrical Power (dBm)
Electrical Power (dBm)
0
~480 ± 5 Hz
-60
-80
-20
-40
~455 ± 5 Hz
0
0
-1
-20 0 20
Delay (μm)
-60
-80
500
1000
1500
0
2000
500
0
(e)
(b)
Electrical Power (dBm)
~1000 ± 5kHz
-40
-60
-80
Intensity (A.U.)
0
-20
1000
1500
2000
Frequency (Hz)
Frequency (Hz)
Electrical Power (dBm)
1
-100
-100
-20
-40
1
0
-1
935 ± 5 Hz
-60
-15 0 15
Delay (μm)
-80
-100
-100
0
500
1000
1500
0
2000
500
Intensity (A.U.)
0
0
(f)
(c)
Electrical Power (dBm)
-20
-40
1000
~1570 ± 5 Hz
-60
-80
500
1000
2000
-20
-40
1
0
-1
-60
1500
2000
-10 0 10
Delay (μm)
-80
~1440 ± 5 Hz
-100
-100
0
1500
Frequency (Hz)
Frequency (Hz)
Electrical Power (dBm)
Intensity (A.U.)
Chapter 3
0
500
1000
1500
2000
Frequency (Hz)
Frequency (Hz)
Fig. 3.8: Doppler frequency of the OCT interference signal observed using the spectrum analyser for
the reference mirror driven at (a) 1 Hz (b) 2 Hz and (c) 3 Hz, and their equivalent Fourier transform
spectrum (d), (e) and (f) using the oscilloscope, respectively. The corresponding interference signal at
their driven frequency is shown in the inset. The Doppler frequency broadening is observed together
with the increment of the driving frequency for both the spectrum analyser’s capture and the
oscilloscope’s Fourier transform spectrum.
36
Chapter 3
3.5 OCT Tomograms
The constructed time-domain OCT system is able to obtain an interference
signal when both the reference and sample arms’ optical path length are matched.
Currently the reference arm consists of a linearly moving mirror secured on a piezo
stage and driven at the rate between 1 to 3 Hz. Lateral scanning was able to move at a
precise manner as the translation actuator has a displacement resolution of 1 µm. Each
lateral movement has been set to 4 or 5 µm, in accordance to the beam spot size.
Using the developed LABVIEW program, the OCT system is operated with the
movement scheme as shown in Fig. 3.9 when scanning a sample.
Lateral scanning.
Scanning optics moves
every 4 or 5 µm per
sequence
Axial
scanning
~1 mm
Cross-section
of an onion
skin
Fig. 3.9: An example of a cross-sectional OCT scan on a highly scattering sample (reproduced from
[3]). The vertical arrows denote depth scanning. The horizontal arrows denote transverse scanning by
moving the scanning optics
Listed in Fig. 3.10 is a preliminary OCT image obtained while scanning a flat
mirror as the sample under test. A mirror is chosen because it is highly reflective and
has even flatness, thus reducing the scattering. Plotted in Fig. 3.10 is the continuous
stack of an envelope function from the interference signal reflected from the mirror
sample. The colour legend in Fig. 3.10 is displayed in a linear scale. The OCT image
of the mirror appears to be skewed instead a linear straight line. The scanning optics
translation using the MM4005 actuator controller is not entirely precise because of the
37
Chapter 3
introduction of error in the actuator displacement. Thus a misaligned movement was
observed during the transverse scanning. Conventional OCT systems sometimes
incorporate signal smoothing or averaging in order to avoid this misalignment defect.
Because this OCT system has a slow scanning rate, at this early stage of development,
this feature will not be embedded into the system until a faster scanning solution to
this has been decided.
Fig. 3.10: OCT image of a mirror sample using the full spectrum of SOA source. Pixel size and
resolution of the image are 200 x 512 and 5 x 0.35 μm in the transverse and axial direction, respectively
The next step in this OCT system study is to introduce a simple biological
sample to be scanned. For this, an onion skin is chosen as the sample under test.
Onion skin is chosen as it has a distinct cell structure that is recognisable (Fig. 3.9).
When using the full bandwidth of the SOA operating at normal driving current, no
interference signal is detected. This might be due to the light signal absorption by the
water content in the onion and the strong scattering of light in the tissue sample. For
this, the full power of the SOA is amplified at higher driving current, which
significantly increases the signal power to + 4.65 dBm. The spectrum of the newly
amplified SOA signal is shown in Fig. 3.11. The centre wavelength of the new
spectrum is shifted to 1525.04 nm. The FWHM bandwidth of the source is slightly
reduced to 57.50 nm. This new spectrum is giving an axial resolution of 17.8 μm. By
increasing the source power, it is hoped that a higher backreflection signal will be
obtained from the onion sample. An interference signal is obtained successfully using
this new source modification.
38
Chapter 3
1.0
Intensity (A.U.)
0.8
FWHM
~57.50 nm
0.6
0.4
0.2
0.0
1400
1450
1500
1550
1600
1650
Wavelength (nm)
Fig. 3.11: The power spectrum of the amplified SOA signal
Fig 3.12(a) and (b) are OCT images of a section of an onion. From the optical
path scan at the reference arm, the images have 175 μm in optical depth. The figures
are then scaled, assuming a sample refractive index of 1.4, which gives a physical
depth of approximately 125 μm. From the OCT images of the onion in Fig 3.11, the
following substructure of the onion skins can be determined. The higher intensity of
red colours denoted the border of the cellular structure. The hollow spaces in the
middle of the cells are denoted by the blue colours.
Unfortunately, the system was not able to properly distinguish the cell
structure in the onion skin as the images appeared to be distorted. Besides the earlier
reason of the actuator displacement error, the limited optical power of + 4.65 dBm is
not powerful enough to penetrate deeper into the sample. The penetration depth of the
light signal too is limited on the onion skin surface at 125 μm. More information of
the cell structure too can be covered if the scanning is performed deeper at 1 to 2 mm.
Nevertheless, the utility of the OCT system was proven although further development
work can be done in order to improve the overall performance of the system.
39
Chapter 3
Fig. 3.12: The OCT image of an onion skin using the amplified spectrum of the SOA source. The
images are scanned at two different positions on the same onion skin. Both images have the pixel size
and resolution of 100 x 512 and 5 x 0.24 μm in the transverse and axial direction, respectively
40
Chapter 3
3.6 Summary
In this chapter, the development and construction of a time-domain OCT
system has been reported. The OCT system was able to distinguish a simple
transparent sample of glass and certain plastic structures. The system has also
successfully stacked and generated a simple OCT image of a mirror using the SOA
broadband source. These experimental results prove the utility of the system to
perform a meaningful, simple sample scan.
The limited backscattering power from the onion skin sample has prevented
the system from scanning as there was no interference signal detected at first, when
the SOA source was used alone. To improve it, the SOA signal was amplified,
increasing the source power using higher driving current. A crude OCT image of an
onion sample skin was obtain. A limited optical scanning depth of 125 μm was one of
the reasons for the distorted OCT image. To compensate for this, the new all fibre
optical delay line will be constructed at the reference arm, which is able to scan at a
deeper penetration. This study is covered in Chapter 5 of this thesis.
41
Chapter 3
3.7 References
1.
J. M. Schmitt, "Optical coherence tomography (OCT): a review," IEEE
Journal of Selected Topics in Quantum Electronics 5, 1205-1215 (1999).
2.
A. M. Rollins, M. D. Kulkarni, S. Yazdanfar, R. Ung-Arunyawee, and J. A.
Izatt, "In vivo video rate optical coherence tomography," Optics Express 3, 219-229
(1998).
3.
W. Dioni, "Homage to the onion skin," in Miscape(2003).
42
Chapter 4
Supercontinuum Source:
Performance Comparison and Noise
Characterisation
4.1 Introduction
The significant difference between OCT and conventional microscopy
is that OCT achieves very high axial image resolutions independent of focusing
conditions, because the axial and transverse resolution are determined independently
by different physical mechanisms. This implies that axial OCT resolution can be
enhanced using a broad bandwidth, low coherence length light source. It is important
to note that the light source not only determines axial OCT resolution via its
bandwidth and central emission wavelength, but also determines both the penetration
in the sample (biological tissue) and the OCT transverse resolution. A source with
high output power with low-amplitude noise is necessary to achieve high sensitivity
and high-speed, real-time OCT imaging. Hence it is obvious that the light source is
the key technological parameter for an OCT system, and a proper choice is essential.
For the OCT case, a very broadband spectrum is required and the selection of the
centre wavelength is crucial to avoid the water absorption peak in biological samples.
For this reason, a supercontinuum source was chosen as it has the better potential to
excel as the future dominant broadband source for use in OCT systems.
A supercontinuum SC450 source unit (see Appendix B), courtesy of Fianium
UK was loaned to this lab for this experimental purpose. Comparing this source with
other commercial supercontinuum lasers [1, 2], this unit is small, compact, light,
robust and allowed ease of handling and mobility despite its inherent features of
Chapter 4
higher spectral density and broader bandwidth. The laser aperture output is built
straight onto the PCF, which is the key ingredient to the nonlinear spectral broadening
supercontinuum generation. Its collimated output provides ease of coupling and stage
mounting (Fig. 4.1 (a)). Other commercial sources mostly have free space laser output,
thus beam steering handling is needed. The setup of commercial Kerr lens modelocked lasers for signal pumping is generally more complex and therefore bulky and
not easy to handle. Whereas for this unit, its passively mode-locked ytterbium (Yb)
fibre laser for picosecond optical pulse pumping is scaled down, comes with very
minimal maintenance and low power consumption. This quasi-cw laser operates at 40
MHz repetition rate. The pump laser was set at the zero-dispersion wavelength of the
PCF. The spectrum of light was broadened when the light propagates through the
fibre. The output spectrum was broadened gradually as the pumping intensity
increased.
Fig. 4.2 (a) reveals the output spectrum of SC450 taken at a portion of its
power. This is done by taking a 4% Fresnel reflection off a glass slide before coupling
it into an SMF. This explains the lower spectral flatness on top of the spectrum, which
is probably due to the cut-off of power at different wavelengths in the fibre. The
supercontinuum light generation was only observed from 600 to 1700 nm, due to the
limited scanning range of the optical spectrum analyser (OSA).
4.2 Broadband Sources Characterisation and Comparison
The OCT imaging requires the usage of a broadband source with a low
coherence property. Using readily available broadband sources in the laboratory, the
loaned supercontinuum unit, SC450, will be compared against semiconductor optical
amplifier (SOA) and erbium-doped fibre amplifier (EDFA) sources. SOA and EDFA
sources are chosen as comparison devices because these two thermal-like sources are
light sources of an incoherent nature where the light packets or photons emitted by
such sources are random in phase when compared against a laser. In addition to that,
these units can provide high power output, more stable and generate a small noise
level in the 1550 nm spectrum.
An SOA is a device which uses semiconductor material as the gain or active
medium [3]. To achieve gain, electrons are excited from the valence band to the
44
Chapter 4
Collimator
output
Metal shielded
PCF
(a)
Glass slide
Beam
dump
Pump
knob
SMF
Thermal
detector
indicator
(b)
Fig. 4.1: (a) The front interface of the supercontinuum SC450 source. (b) Because the average power
output from SC450 is too high, coupling of its collimated beam into SMF has to be done by taking a
partial light beam reflected off a glass slide.
45
Chapter 4
conduction band with electrical current. In the absence of an input signal, a certain
fraction of the excited molecules or atoms will fall back to the low energy level and
produce spontaneous emission photons. These photons emit in random directions and
a small fraction of these photons will propagate towards the output of the amplifier.
As the photons travel along the waveguide towards the output, these photons will
stimulate the generation of other photons. The stimulated photons will have the same
direction of propagation, frequency, phase, and polarization as the photons that caused
the stimulated emission. These amplifiers have a similar structure to Fabry Perot laser
diodes but with anti-reflection design elements at their end faces. SOA has the
potential for practical use in optical networking due to advances in optical
semiconductor fabrication techniques and design, particularly in short-haul optical
transmission.
Rare-earth doped optically pumped fibres too are normally used in optical
communication to amplify signals. Erbium, for example, can realize the operation
around the 1550 nm wavelength, the C-band in optical communication. EDFAs have
two commonly-used pumping bands – 980 nm and 1480 nm. The 980 nm band has a
higher absorption cross-section and is generally used where low-noise performance is
required. The absorption band is relatively narrow and so wavelength stabilised laser
sources are typically needed. The 1480 nm band instead has a lower, but broader,
absorption cross-section. In this study, the EDFA with a 980 nm pump wavelength is
used. The 980 nm pump light is multiplexed into the erbium doped fibre where the
optical signal is amplified through interaction with the doping erbium ions via the
ASE process.
ASE itself is basically an amplified “noise” [3]. An EDFA can amplify not
only the input field from a laser but also the spontaneous radiation emitted by the
excited molecules or atoms of the amplifier itself. The EDFA is actually bidirectional,
with ASE leaving from both ends. Spontaneous emission occurs over the entire length
of the optical amplifier. A spontaneous emission photon that travels the whole length
of the amplifier can stimulate the emission of more photons and lead to substantial
output radiation at the other end of the amplifier. ASE is a “noise” [3] because it can
significantly deplete the upper level population, thus diminishing the gain available to
the input signal to be amplified. ASE can be expected to seriously deplete the upperlevel population of the amplifying transition if it becomes comparable in magnitude to
the saturation intensity of the transition.
46
Chapter 4
-10
(a)
1.0
-30
Intensity (A.U.)
Power (dBm)
-20
-40
-50
-60
-70
(d)
FWHM
~9.6 ± 0.1 µm
0.5
0.0
-0.5
-1.0
-80
600
800
1000
1200
1400
1600
-100 -80 -60 -40 -20
Wavelength (nm)
-10
20
40
60
80
Delay (µm)
(b)
1.0
-20
-30
Intensity (A.U.)
Power (dBm)
0
-40
-50
-60
-70
(e)
FWHM
~17.6 ± 0.1 µm
0.5
0.0
-0.5
-1.0
-80
1400
1450
1500
1550
1600
1650
1700
-100 -80 -60 -40 -20
Wavelength (nm)
-10
(c)
1.0
Intensity (A.U.)
Power (dBm)
-40
-50
-60
20
40
60
80
Delay (µm)
-20
-30
0
-70
(f)
FWHM
~32.3 ± 0.1 µm
0.5
0.0
-0.5
-1.0
-80
-100 -80 -60 -40 -20
1480 1500 1520 1540 1560 1580 1600 1620
0
20
40
60
80
Delay (µm)
Wavelength (nm)
Fig. 4.2: Full spectrum of (a) SC450 (b) SOA (c) EDFA sources and their respective interference signal
when launched into the OCT setup. The 3 dB bandwidths of each source are 1440.0 nm (see Appendix
B for characteristics) for the SC450, 59.1 nm for the SOA and 37.0 nm for the EDFA. The optical
spectrums of these sources are recorded with a 0.5 nm resolution bandwidth.
47
Chapter 4
Because of the undepleted pump, the full spectrum of the SC450 in Fig. 4.2 (a)
exhibited an obvious peak at 1064 nm. This corresponds to pumping by the modelocked Yb fibre laser at the zero-dispersion wavelength of the nonlinear PCF in the
SC450. No pump peak appearance is observed for the EDFA at 980 nm because the
unit is backward pumped and the ASE output is observed at its front end output. It is
also noticed that a second peak appears at 766 nm for the SC450. This might be
contributed from the nonlinear interaction generated in the PCF. The exact nonlinear
effects that contributed to this second peak are currently unknown. Generally
speaking, χ(2) nonlinear process does not appear in silica, a centrosymmetric material.
The 1064 nm photon which is clearly distinct in this source spectrum may have
interacted together along with the nonlinear process to produce this peak.
48
Chapter 4
4.2.1 Interference Signal Observation
To demonstrate the application of broadband supercontinuum light from
SC450 for high-resolution OCT, its interference signal was obtained and measured
against the equivalent interference signal of the SOA and EDFA source. The full
spectrum of each source was launched into the OCT setup to obtain their interference
signal. Because of the wavelength cut-off and loss each time the supercontinuum
spectrum traveled in the OCT setup, the final spectrum incident on the photodetector
will be spectrally shaped, when compared to the SC450 original spectrum, as shown
in Fig. 4.7. This loss is mostly attributed to the 3 dB coupler cut-off. From the 3dB
bandwidth, Δλ in Fig. 4.7, using Eq. 2.14, the axial resolution for the SC450 is
calculated to be 4.8 μm. The measured FWHM of the interference signal in Fig. 4.2(d)
is 9.6 μm, twice the magnitude of the theoretical value. The reason for the fringe
broadening is because the SC450 spectrum, at this cut-off wavelength, does not
exhibit partial coherence features, a prominent aspect of a thermal-like source.
Similarly to the one reported in the earlier Section 3.3, the interference signal for the
SOA source (Fig. 4.2 (b)) is measured to be 17.5 μm, matching its theoretical value.
Because the SOA full spectrum resembled a Gaussian-like shape, its interference
signal too fits a Gaussian-like shape. The interference signal for the EDFA source, in
Fig. 4.2 (f), resembles a sinc function shape, where side lobes appeared along side the
main lobe. The reason for this side lobes appearance is because the full spectrum of
the EDFA resembles a flat top shape. The FWHM recorded for the main lobe is 32.3
μm. This value does not correspond with the theoretical FWHM value of Gaussian
shape spectrum (Eq. 2.14), but instead fits well with the general sinc shape equation,
given by Eq. 4.2.
Fig. 4.3 shows the experimental setup that is used to filter the broadband
sources by taking the reflectivity profile of a linearly chirped fibre Bragg grating
(CFBG). Here the effects of shaping the sources’ spectra are compared with their
obtained interference signal. Two CFBG are used, i.e.: the flat top and Gaussian
shaped filters. The broadband source is optically filtered with the CFBG using the 3
port circulator configuration. The flat top grating is 140 mm long and reflecting ~40
nm bandwidth at the centre Bragg wavelength of 1550 nm with 63% reflectivity. The
Gaussian grating is 130 mm long and reflecting ~23 nm bandwidth at the centre
wavelength of 1545 nm with 70% reflectivity. The end of the CFBG is immersed in
49
Chapter 4
CFBG
λshort
Index matching
fluid
λlong
Broadband
source
OSA
Optical
attenuator
Fig. 4.3: Schematic diagram of signal shaping setup using CFBG for spectral and interference signal
characterization.
index matching fluid to prevent the backreflection signal. For all sources, optical
power of -15 dB was maintained, using the 1550 nm optical attenuator, before feeding
the signal into the OSA. The filtered spectrum for each broadband source is listed in
Fig. 4.5 and Fig. 4.6.
Then, modifying the experimental setup in Fig. 4.3, the filtered broadband
signals are reused by launching them into the OCT setup, as depicted in Fig. 4.4, to
obtain their interference signal. Using the two earlier CFBGs with different shaping,
their interference signals using these three broadband sources was characterized, as
listed in Fig 4.5 and Fig. 4.6.
λshort
Broadband
source
CFBG
Index matching
fluid
λlong
PC
C
M
OBJ
M
Reference
arm
3 port
circulator
3dB coupler
C
PD
Oscilloscope
Sample
arm
Fig. 4.4: Interference signal characterization of the broadband sources using the two different optically
filtered spectral shapes. The OCT system is highlighted in the shaded box.
50
Chapter 4
-40
1.0
-50
(d)
FWHM
~38.9 ± 0.1 µm
-52
-60
1540 1545 1550
-70
-80
-90
Intensity (A.U.)
Power (dBm)
-50
-48
(a)
-100
0.5
0.0
-0.5
Width
~60.0 ± 0.1 µm
-1.0
-110
1480 1500 1520 1540 1560 1580 1600 1620
-100 -80 -60 -40 -20
Wavelength (nm)
0
20
40
60
80
40
60
80
40
60
80
Delay (µm)
-40
-60
(b)
1.0
Δλ = 40 nm
-70
-80
-90
Intensity (A.U.)
Power (dBm)
-50
-100
(e)
0.5
0.0
-0.5
-1.0
-110
1480 1500 1520 1540 1560 1580 1600 1620
-100 -80 -60 -40 -20
Wavelength (nm)
0
20
Delay (µm)
-40
(c)
1.0
-60
Intensity (A.U.)
Power (dBm)
-50
-70
-80
-90
-100
(f)
0.5
0.0
-0.5
-1.0
-110
-100 -80 -60 -40 -20
1480 1500 1520 1540 1560 1580 1600 1620
0
20
Delay (µm)
Wavelength (nm)
Fig. 4.5: The filtered spectra of (a) SC450 (b) SOA and (c) EDFA sources using a flat top shaped
CFBG and their respective sinc shaped function interference signals. Inset: (a) Enlarged spectral
modulation on top of the top hat spectrum. (c) Small arrow denotes the small dip in spectrum. The
filtered optical spectrums are from the OSA with a 0.01 nm resolution bandwidth.
51
Chapter 4
-40
-60
(a)
Δλ = 23 nm
-70
-80
-90
-100
FWHM~
51.0 ± 0.1 µm
(d)
1.0
Intensity (A.U.)
Power (dBm)
-50
0.5
0.0
-0.5
-1.0
-110
1480 1500 1520 1540 1560 1580 1600 1620
-80
-60
-40
Wavelength (nm)
-20
0
20
40
60
80
40
60
80
40
60
80
Delay (µm)
-40
(b)
(e)
1.0
-60
Intensity (A.U.)
Power (dBm)
-50
-70
-80
-90
-100
0.5
0.0
-0.5
-1.0
-110
1480 1500 1520 1540 1560 1580 1600 1620
-80
-60
-40
Wavelength (nm)
-20
0
20
Delay (µm)
-40
(c)
(f)
1.0
-60
Intensity (A.U.)
Power (dBm)
-50
-70
-80
-90
-100
0.5
0.0
-0.5
-1.0
-110
-80
1480 1500 1520 1540 1560 1580 1600 1620
-60
-40
-20
0
20
Delay (µm)
Wavelength (nm)
Fig. 4.6: The filtered spectra of (a) SC450 (b) SOA and (c) EDFA sources using a Gaussian shaped
CFBG and their respective Gaussian shaped function interference signals. The filtered optical
spectrums are from the OSA with a 0.01 nm resolution bandwidth.
52
Chapter 4
-35
-40
Δλ = 211.1 nm
Power (dBm)
-45
-50
-55
-60
-65
-70
600
800
1000
1200
1400
1600
Wavelength (nm)
Fig. 4.7: The SC450 spectrum that is incident on the photodetector after coupling from the OCT
coupler. This optical spectrum is recorded with a 0.5 nm resolution bandwidth.
For the flat top filtered SC450 spectrum, there is a periodic wavelength
oscillation in this 1550 nm wavelength region, as shown in the inset of Fig. 4.5 (a).
The SC450 source is emitting a very broad bandwidth of more than 1300 nm. Hence
there will be interference of modes when it is coupled into optical fibre from its
collimator output. The induced intermodal interference will cause mode competition
at the end of the fibre where two or more wavelengths may be experiencing
constructive interference [4]. To prevent this, a better beam coupling method was
used to ensure that a proper beam profile was coupled into the fibre core. Therefore,
no mode interference is observed in Fig. 4.6 (a) during the Gaussian shaped filter
experiment.
For both the flat top and the Gaussian filtered SC450 spectrum, the FWHM of
their interference signals, were 38.9 μm and 51.0 μm respectively. The same values
were obtained for SOA and EDFA filtered sources. Due to uneven fringe symmetry,
the correct FWHM measurements for Gaussian filtered sources were obtained from
the envelope function of their interference signal in Fig. 4.6. For the flat top filtered
53
Chapter 4
spectrum, the width of its main lobe is 60.0 μm. At their equivalent time value, the
main lobe width and the FWHM of the sinc-shaped interference signal conform to the
sinc-shape general equation given by Eq. 4.2. Bigger sidelobes appeared for the
EDFA filtered interference signal (Fig. 4.5 (f) and Fig. 4.6 (f)) because of a small dip
in the spectral shape at the 1535 nm region which moderately shaped the spectrum.
The theoretical axial resolution for the Gaussian shaped spectrum is 49.0 μm, which
does not differ much from the measured FWHM of the interference signal at 51.0 μm.
The Gaussian shaped spectra conform to their theoretical axial resolution, as Eq. 2.14
can only be applied to a Gaussian shaped spectrum.
54
Chapter 4
4.2.2 Effects of Optical Filters Shaping to Its Coherence Function
The sinc shaped functions in Fig. 4.5 (d), (e) and (f) are actually the Fourier
transform of the filtered spectrum in Fig. 4.5 (a), (b) and (c), respectively. The
common sinc function; sinc(θ) = sin (πθ) / πθ satisfies the interference signal shape in
Fig. 4.5 (d), (e) and (f).
The Fourier transform of a rectangular pulse in Fig. 4.8 (a) is a sinc function
(Fig. 4.8 (b)) as denoted by the following relationship
⎧1,
x1 (t) = ⎨
⎩0,
t < T1
t > T1
T1
X1 (ω) = ∫ e − jωt dt = 2
←⎯→
F
−T1
sin ωT1
ω
.
(4.1)
Now consider the signal with the Fourier transform in Fig. 4.8 (d) and its inverse (Fig.
4.8 (c)), as revealed in the following relationship
⎧1,
X 2 (ω) = ⎨
⎩0,
ω <W
ω >W
W
F
←⎯→
x 2 (t) =
∫e
jϖt
−W
dω =
sin Wt
.
πt
(a)
(b)
(c)
(d)
Fig. 4.8: Duality property of Fourier transform.
55
(4.2)
Chapter 4
This is the consequence of the duality property of the Fourier transform.
By comparing the Fourier transform and its inverse transform relations,
x(t ) =
1
2π
∫
+∞
−∞
X (ω )e jϖt dω
(4.3)
+∞
X (ω ) = ∫ x(t )e − jϖt dt
(4.4)
−∞
it is noticed that the equations are similar in form but not quite identical. This
symmetry was also noticed in Eq. 4.1 and Eq. 4.2, i.e. the transform of the flat top
function was the sinc function and vice versa.
A Gaussian is an example of a self reciprocal function, in other words, both
the function and its transform have the same form (Fig. 4.9)
∞
x(t) = exp(−πt 2 )
X(ω) = ∫ e ( −πt ) e − jϖt dt = exp(−πω 2 ) .
F
←⎯→
2
−∞
x(t)
(4.5)
X(ω)
F
ω
t
Fig. 4.9: Fourier transform of a Gaussian.
In OCT applications, a Gaussian signal is preferred in order to prevent the sidelobes
appearing in the interference signal. A Fourier transform of a Gaussian is another
Gaussian and vice versa. Thus using a Gaussian-like spectrum from the broadband
source, the interference signal will be a Gaussian too.
56
Chapter 4
4.3 Relative Intensity Noise Measurement
Unfortunately, the supercontinuum unit, SC450 suffers from noise beating,
largely contributed from its intensity noise. The noise associated with supercontinuum
generation, depending on the input pulse parameters, can lead to 50% intensity
fluctuations. As pointed out by Corwin et al. [5] and Newbury et al. [6], the origin of
this noise is the nonlinear amplification of quantum fluctuations, both in the input
laser light and in the Raman scattering process within the PCF. This noise is attributed
to two distinct components: low frequency (<1 MHz) noise in the supercontinuum
generation is an amplified version of the amplitude noise of the pump source and a
broadband component originating from nonlinear amplification of quantum
fluctuations in both the input laser light and in the Raman scattered light. The Raman
scattering here only plays a relatively minor role as compared to input shot noise from
the pump. Numerically these authors simulated the above noise contributions using
the generalised nonlinear Schrödinger equation and found that when only Raman
scattering is included, the RIN is reduced by ~20 dB. But, when only the shot-noise is
included, the RIN is reduced by less than ~1 dB. For this reason, it is interesting to
study the contribution of the shot-noise from the pump towards the intensity noise of
SC450 supercontinuum source.
Generally, noise is known to vary as a function of optical intensity, electrical
frequency, chromatic dispersion, the source bandwidth and its spectral shape that is
incident on the photodetector [7]. The noise incident on the photodetector normally
results from the shot, thermal and intensity noise components.
Shot noise is produced by the random nature of photons arriving at the
detector. For the PIN photodiode, the shot noise power, in a one Hz bandwidth, is
given by
N shot = 2qI dc R L .
Here Idc is the average current, q and RL are the electron charge and detector load
resistance respectively.
57
(4.6)
Chapter 4
Thermal noise on the other hand, is produced by the random thermal motion of
electrons in the load resistor. The existence of this noise can limit the sensitivity of the
detector. Its power spectral density (PSD), Nth can be described by
N th = 4k B T
(4.7)
where kB is the Boltzmann constant and T is the absolute temperature.
B
The total system noise, NT at the detector output is the summation of these two
noise sources and the source intensity noise, Nsource
NT = Nsource + Nshot + Nth.
(4.8)
The noise of interest in this study is the intensity noise, Nsource, which refers to the
noise generated by the source itself. This noise is caused by the intensity fluctuations
due to the laser’s structural parameters such as the physical nature of the PCF or its
pump laser in the SC450 case.
The noise of a broadband source can also be parameterized as the relative
intensity noise (RIN) on the photodetector. RIN measurement describes the source’s
maximum available amplitude range for signal modulation. It also serves as a quality
indicator of laser devices. RIN is the ratio of noise power spectral density or the
mean-square optical intensity noise ΔP 2 fluctuation (in a one Hz bandwidth) to the
square of the total average power, P
RIN =
ΔP 2
P
2
Hz-1.
(4.9)
RIN is essentially a dynamic range measurement. If the average power is
reduced, the RIN measurement range is reduced. Thus the thermal noise limitation
can be overcome when the average power is large enough.
58
Chapter 4
λshort
CFBG
Index matching
fluid
λlong
Broadband
source
Optical
attenuator
HP
Lightwave
Analyzer
Fig. 4.10: Experimental setup for RIN measurement.
In order to characterize the intensity noise of SC450, the experimental setup
presented in Fig. 4.10 was constructed. At the same time the noise intensity of SOA
and EDFA sources too are measured. These two light sources are known to be stable
and, most importantly, suit the low noise criteria for comparison with the SC450.
These two units were chosen to be the benchmark for comparison with SC450. The
optical signal from each of the broadband sources was filtered before launching into
the lightwave analyser (HP71400C, Agilent). Two different CFBG filters were used,
i.e.: flat top and Gaussian shaped filter, similar to the ones reported in the earlier
Section 4.2.1. Thus the RIN values of these sources were compared as a function of
their spectral shape. An optical attenuator was used to maintain the same -15 dBm
optical power into the lightwave analyser’s internal, built-in, photodetector during the
measurement. The lightwave analyser has a 22 GHz bandwidth. The resolution and
video bandwidth of the lightwave analyser were set at 1 MHz and 100 kHz
respectively. The electrical signal attenuation is set at 0 dB and the measurements are
subject to 32 times sample averaging to smooth out the result reading. The trace data
is obtained in the 1 MHz – 1 GHz frequency regime. The trace data downloaded from
the lightwave analyser of each filtered sources is summarised in Table 4.1.
A quick check using the HP polarisation analyser (HP 8509B, Agilent)
revealed that the Gaussian filtered SC450 source is polarised with the degree of
polarisation of 82.5 %, when the filtered spectrum is launched into the unit. The full
bandwidth unpolarised SC450 light suffers induced polarisation effects upon filtering
and cut-off of wavelength, together with loss, upon incidence on the built-in
photodetector in the polarisation analyser. This high degree of polarisation is known
to introduce at least a factor of 2 in the RIN measurement of a light source [8],
compared with that of unpolarised light. For polarised light, the electric field vector of
the light’s electromagnetic field can be arbitrarily divided into two perpendicular
59
Chapter 4
components, where both are traveling in phase along its direction of propagation.
Hence most of the energy will be confined in this vector. This thus gives rise to the
factor of 2 increase in the intensity noise of the light source. No polarisation effect
was observed for SOA and EDFA.
For comparison, the measured and predicted value of RIN for each of the
filtered sources is summarized in Table 4.2. Recall that RIN value is the ratio of the
noise power spectral density (PSD) of the spectrum to the square of the average power.
For unpolarised thermal light, the RIN at the desired frequency, f , of interest, is
given by this expression [8]:
R( f ) = ∫
∞
0
S (υ ) S (υ + f )
dυ
P2
(4.10)
where S(υ) is the noise PSD as measured by the OSA and P is the total optical power
in the spectrum. From Eq. 4.10, a MATLAB program was written to evaluate each
source’s RIN value and the results were compared with the measured RIN value. The
integral is implemented in a summation model. This approximation is only true
provided that the light source is a true thermal light source. With help from Dr.
Anoma McCoy with the modelling, the RIN value for the filtered SOA and EDFA
spectrum was predicted and is quoted in Table 4.2.
For this evaluation, 500 MHz was chosen because, at this frequency, the
marker point in the screen dump stays safely at the signal base and, for the SC450
case, it does not overlap at the spike peak which later can contribute to a wrong RIN
measurement. The presence of 1/f noise in the source output can impose a maximum
achievable signal-to-noise ratio in detection sensitivity. In a wide range of carrier
transfer processes, the power spectrum, which is inversely proportional to frequency,
diverges at low frequencies [9]. This is observed in the noise spectrum where the lowfrequency end will exhibit a different slope. The selection of the 500 MHz marker too
is influenced by this condition. It is selected because it is positioned far away from the
low-frequency end.
The RIN values for the SC450, SOA and EDFA filtered signal are obtained
from the “RIN (Laser)” caption in their respective screen dumps in Table 4.1 and are
summarised in Table 4.2. The equivalent RMS power is listed as figures enclosed in
brackets in Table 4.2. From the Table 4.2, the measured RIN values for the SOA and
60
Chapter 4
EDFA sources filtered with a flat top filter are in good agreement with the predicted
RIN value. Their measured RIN values have repeatability within ± 0.25 dB tolerance
each time the measurement is performed. For the flat top filtered source, the RIN
value for the EDFA source is slightly higher than for the SOA source because of a
small dip in the spectral shape at the 1535 nm region (Fig. 4.5 (c)). The opposite was
observed for the Gaussian filtered source. The RIN value for the SOA is instead
higher because the FWHM bandwidth of its Gaussian filtered shape is smaller than
the FWHM bandwidth of the Gaussian filtered EDFA source (Fig. 4.6). The RIN
values for light sources are generally higher when measured for the Gaussian spectral
shape because of their steep roll over [7] as compared to the flat top spectral shape.
61
Chapter 4
Table 4.1: RIN measurement on a HP71400C Lightwave Analyser for SC450, SOA
and EDFA sources in the 1MHz – 1GHz frequency regime
Gaussian filter (23 nm bandwidth)
-10
-10
-20
-20
-30
MKR FRQ 500 MHz
-61.14 dBm (1 Hz)
-40
-50
-60
RIN (Laser)
= -95.30 dB/Hz
RIN (System)
= - 95.29 dB/Hz
Thermal Noise Term = -120.61 dB/Hz
Shot Noise Term
Optical power (dBm)
SC450
Flat top filter (40 nm bandwidth)
Optical power (dBm)
Source
-30
-50
-60
= -138.63 dB/Hz
-70
-70
200
400
600
800
MKR FRQ 500 MHz
-62.36 dBm (1 Hz)
-40
1000
RIN (Laser)
Shot Noise Term
200
-10
-10
-20
-20
-30
-50
MKR FRQ 500 MHz
-73.19 dBm (1 Hz)
-60
RIN (Laser)
= -125.84 dB/Hz
RIN (System)
= -119.39 dB/Hz
Thermal Noise Term = -120.57 dB/Hz
Shot Noise Term
-50
-60
RIN (Laser)
= -124.62 dB/Hz
RIN (System)
= - 119.15 dB/Hz
Thermal Noise Term = -120.67 dB/Hz
Shot Noise Term
400
600
800
1000
200
= -138.66 dB/Hz
-10
-10
-20
-20
-30
MKR FRQ 500 MHz
-73.10 dBm (1 Hz)
-40
RIN (Laser)
= -124.80 dB/Hz
RIN (System)
= - 119.23 dB/Hz
Thermal Noise Term = -120.71 dB/Hz
Shot Noise Term
400
600
800
1000
Frequency (MHz)
Optical power (dBm)
Optical power (dBm)
1000
MKR FRQ 500 MHz
-73.05 dBm (1 Hz)
-40
Frequency (MHz)
-60
800
-70
200
-50
600
-30
= -138.63 dB/Hz
-70
EDFA
400
= -138.66 dB/Hz
Frequency (MHz)
Optical power (dBm)
SOA
Optical power (dBm)
Frequency (MHz)
-40
= -97.80 dB/Hz
RIN (System)
= - 97.78 dB/Hz
Thermal Noise Term = -120.60 dB/Hz
-30
MKR FRQ 500 MHz
-73.10 dBm (1 Hz)
-40
-50
-60
= -138.65 dB/Hz
RIN (Laser)
Shot Noise Term
-70
= -125.03 dB/Hz
RIN (System)
= - 119.24 dB/Hz
Thermal Noise Term = -120.64 dB/Hz
= -138.65 dB/Hz
-70
200
400
600
800
1000
Frequency (MHz)
200
400
600
Frequency (MHz)
62
800
1000
Chapter 4
Table 4.2: Measured and predicted RIN for SC450, SOA and EDFA sources
Flat top filter (dB/Hz)
Source
SC450
SOA
EDFA
Measured
Predicted
-95.30
(2.9512 x 10-10 Hz-1)
-125.84
(2.6062 x 10
-13
Gaussian filter (dB/Hz)
-1
Hz )
-124.80
(3.3113 x 10-13 Hz-1)
N.A.
-126.6
-126.2
Measured
-97.80
(1.6596 x 10-10 Hz-1)
-124.62
(3.4514 x 10-13 Hz-1)
-125.03
(3.1405 x 10-13 Hz-1)
Predicted
N.A.
-125.31
-125.56
For the SC450 source, it is noticed that a spike peak is observed at every 40
MHz, in Table 4.1, which corresponds to the repetition rate of this pulsed pumped
broadband source. Take note that the RIN value measured for SC450 was higher in
magnitude compared to the SOA and EDFA, for both the flat top and Gaussian filter.
The flat top filtered SC450 produced a power fluctuation of ~30.54 dB and ~29.50 dB
times higher than the SOA and EDFA, respectively. For the Gaussian filtered
spectrum, SC450 produced a power fluctuation of ~26.82 dB and ~27.23 dB times
higher than the SOA and EDFA, respectively. This increase resulted from two
combining factors. In addition to the earlier factor of 2 increment in intensity noise
from a polarised signal, the nonlinear amplification of the intensity noise in the SC450
input pumping laser, due to the complex nonlinear phenomena that include SPM,
FWM and SRS, are contributing to this higher RIN reading. Unfortunately the
measured RIN values for SC450 are not in the vicinity of the thermal-like source’s
predicted value, hence ‘N.A’ is written in Table 4.2. This can be attributed to the same
reasons as mentioned before. The predicted RIN value can only be achieved if the
source is a purely thermal-like source. Thus, it can be concluded that SC450 is not a
thermal light source.
Nevertheless, in terms of the OCT application, the source can still be utilized
as a potential candidate. Even though, in terms of power variation, SC450 seems to be
noisier, its interference signal in Fig. 4.2 (d), Fig. 4.5 (d) and Fig. 4.6 (d) appeared to
be stable enough to provide good axial resolution for OCT applications.
63
Chapter 4
4.4 Summary
The utility and usability of a SC450 supercontinuum source for OCT
applications has been covered in this chapter. The interference signal of the SC450
has been compared with the interference signal of the SOA and EDFA. Their fringe
shapes were compared at their full spectrum, flat top and Gaussian filtered spectrum.
The filtered shapes correspond well with their Fourier transform shapes at their
interference signal; flat top will give a sinc shape function while Gaussian will give a
Gaussian shape function. At their Gaussian filtered shape, the interference signals
conform well with the coherence length theoretical values.
The RIN values for the broadband sources have been compared too. Again, the
RIN value has been compared for their flat top and Gaussian filtered spectrum.
Generally, SC450 gives a higher RIN value; hence it is noisier when compared with
the SOA’s and EDFA’s. At their flat top filtered spectrum, the EDFA’s RIN value is
higher than SOA’s. But for Gaussian filtered spectrum, the SOA’s RIN value is higher
than EDFA’s. The predicted RIN value for both flat top and Gaussian filtered, SOA
and EDFA sources matched well with their measured RIN value.
It can be concluded that, even though the SC450 source is noisier in terms of
power fluctuation or RIN value, when compared with SOA’s and EDFA’s, at least its
source signal showed a stable shape throughout the characterization of its interference
signal. This at least can guarantee a very functional usage for OCT applications.
64
Chapter 4
4.5 References
1.
http://www.koheras.com.
2.
http://www.precisionphotonics.com.
3.
D. Derickson, Fiber Optic Test and Measurement (Prentice Hall PTR, Upper
Saddle River, New Jersey, 1998).
4.
I. Turek, I. Martincek, and R. Stransky, "Interference of modes in optical
fibers," Optical Engineering 39, 1304-1309 (2000).
5.
K. L. Corwin, N. R. Newbury, J. M. Dudley, S. Coen, S. A. Diddams, B. R.
Washburn, K. Weber, and R. S. Windeler, "Fundamental amplitude noise limitations
to supercontinuum spectra generated in a microstructured fiber," Applied Physics B:
Lasers and Optics 77, 269-277 (2003).
6.
N. R. Newbury, B. R. Washburn, K. L. Corwin, and R. S. Windeler, "Noise
amplification during supercontinuum generation in microstructure fiber," Optics
Letters 28, 944-946 (2003).
7.
A. D. McCoy, "Intensity Noise Suppression Using a Semiconductor Optical
Amplifier: Characterisations and Applications," in Ph.D Thesis, Optoelectronics
Research Centre(University of Southampton, Southampton, 2005).
8.
G. E. Obarski, and J. D. Splett, "Transfer standard for the spectral density of
relative intensity noise of optical fiber sources near 1550 nm," Journal of the Optical
Society of America B: Optical Physics 18, 750-761 (2001).
9.
M. S. Keshner, "1/f NOISE," Proceedings of the IEEE 70, 212-218 (1982).
65
Chapter 5
All Fibre Optical Delay Line for
Optical Coherence Tomography
Application
This chapter will review the technologies and techniques behind the
construction of an all fibre optical delay line. First, the types of optical delay lines
used in the OCT system will be briefly presented. Overview of the CFBG used in the
construction of an all fibre optical delay line will be given as well. Then, the
development and construction of the all fibre optical delay line will be reported.
5.1 Introduction to Optical Delay Line
Scanning range and repetition rate are one of the major parameters in optical
delay line. In most applications, it is desirable to have a high repetition rate and large
scanning range to improve the signal-to-noise ratio (SNR) of the application system.
Among other applications, that use the optical delay line as its key component, are
[1]:
1. Frequency-resolved optical gating (FROG) for the measurement and
characterization of ultrafast pulses.
2. Optical autocorrelation for the measurement of intensity versus time of
optical pulses.
3. Clocking and delay generation for time-division multiplexing in optical
communications.
4. Optical low coherence reflectometry (OLCR) for fault detection in optical
waveguides.
Chapter 5
In the time-domain OCT setup, optical delay line techniques based on linear
translation of reflective elements, rotational methods, fibre stretching and group delay
generation based on Fourier domain technique have been reported [2]. In this chapter,
the delay line technique based on FBG will be reviewed and reported. This technique,
when compared with the previous techniques, can provide greater mobility from the
elimination of bulk optics, a larger path delay amplification and subsequently deeper
depth penetration into sample.
5.2 Overview of Fibre Bragg Grating
Fibre Bragg gratings (FBGs) are typically made in silica optical fibres with
cores doped with germanium. These germanium-doped fibres will exhibit increased
photosensitivity and, together with the introduction of hydrogen into the core, this will
further enhance the photosensitivity effect. These two effects combine to create the
periodic variation of the refractive index in the core region when subject to a periodic
pattern of UV illumination.
The common fabrication technique of FBGs nowadays is using a phase mask
[3]. Upon the incidence of UV beam, the grooves on the UV-transmissive mask plate
will cause the incidence beam to be diffracted, into several orders, m = 0, ±1, ±2….
This is shown schematically in Fig. 5.1.
Incident
UV beam
θi
Λpm
Phase mask
m = -1
θm/2
θt
Transmitted beam, m = 0
Fig. 5.1: Diffraction of UV beam upon incidence on a phase mask.
67
Chapter 5
The incident and diffracted orders satisfy the general diffraction equation, with the
period, Λpm, of the phase mask
Λ pm =
mλUV
⎛ θm
⎞
− sin θ i ⎟
⎜ sin
2
⎝
⎠
(5.1)
where θm/2 is the angle of the diffracted order, λUV is the wavelength, and θi is the UV
beam’s incident angle. Normally, the UV beam is incident normal to the phase mask
and most of the diffracted light is contained in the zero, +1, and -1 diffracting orders.
In a uniform FBG, different wavelength components of an incident light are
reflected at different positions along the grating, according to the local grating period
[3]
λ B = 2neff Λ
(5.2)
where λB is the Bragg wavelength, neff is the effective refractive index of the core
B
mode and Λ is the grating period.
Most gratings designed for practical applications are non uniform gratings.
Often the main reason is to reduce the undesirable sidelobes. While a uniform FBG
has the same grating period along its whole length, a chirped fibre Bragg grating
(CFBG) has a grating period which varies along its length which can take any
functional form of linear, quadratic or even higher order. Chirping the period of the
grating enables the dispersive properties of the scattered light to be tailored. CFBGs
are useful for dispersion compensation, for controlling and shaping shorter pulses in
fibre lasers and for creating stable continuous-wave (cw) and tunable mode locked
external-cavity semiconductor lasers [4]. Linearly chirped FBGs are popular due to
their linear group delay properties that can be used for performing dispersion
compensation. Owing to the varying period, the spectrum of the CFBG is broader
compared to a uniform FBG.
Fig. 5.2 shows a schematic of a chirped grating, of length Lgr and chirped
bandwidth ΔΛchirp. Note that the chirp in the period can be related to the chirped
bandwidth, ΔΛchirp of the fiber grating as
68
Chapter 5
Δλchirp = 2neff(Λlong – Λshort)
(5.3)
= 2neff(ΔΛchirp).
The reflection from a chirped grating is a function of wavelength and, therefore,
broadband light entering into a positively chirped grating (increasing period from
input end) will see the shorter wavelength component, λshort reflected earlier than the
longer one, λlong.
Λshort
Λ0
Λlong
Lgr
Fig. 5.2: Chirped grating.
Chirped gratings disperse light by introducing a delay between the shortest and
longest reflected wavelengths. This dispersion can be used to compensate for
chromatic dispersion induced broadening during optical transmission in a fibre.
Many OCT systems rely on bulk optics for an optical delay line in their
reference arm. An all-fibre optical delay line in an OCT system can be made possible
by constructing two CFBGs and positioning them in reversed cascaded orientation
against each other. The idea of all fibre optical delay line for OCT applications was
first proposed by Lee et al. [5]. The authors used a pair of CFBGs, and cascaded them
in a reverse orientation manner to produce a delay when one of the gratings was
stretched. By stretching one of the gratings, an optical delay of a few orders of
magnitude greater than the actual physical stretch can be achieved. Much of the
theoretical treatment for this method has been dealt with rigorously by Lee et al.[5]
and Yang et al.[6]. Using this method, all fibre material can be realised in this
mechanism where the inherent fibre advantage of low insertion loss, alignment free
and ease of mobility can be achieved. The all fibre delay line too can offer amplified
optical delay for the scanning of samples in the OCT system.
While straining the fibre under tension, it will introduce a wavelength shift in
the grating. This shift of Bragg wavelength with strain can be expressed using [7]
69
Chapter 5
ΔλB = λB (1 − pe )ε
(5.4)
where ε = dl/Lgr is the applied strain, pe is the elasto-optic coefficient, with numerical
value ~0.22. For a chirped dispersion grating, its dispersion D is given by
D=
2n ⎛⎜ Lgr
⋅
c ⎜⎝ Δλchirp
⎞
⎟
⎟
⎠
(5.5)
where n is the effective refractive index, Lgr is the fibre length, c is the light speed and
Δλchirp is the chirped grating’s bandwidth. The dispersion D is also associated with the
group delay tg(λ) as
D (λ ) =
dt g (λ )
dλ
(5.6)
where the group delay tg(λ) depends on the wavelength.
The presence of the time or phase ripple in the CFBGs group delay has been
thought to contribute to the significant broadening in the interference signal in an
OCT system [8]. Through the group delay, the phase of the light signal changes while
travelling along the fibre. The group delay, tg, experienced by the light pulse, is
related to this phase shift, φ , by:
t g (λ ) =
dφ (ω )
λ2 dφ (λ )
=−
⋅
dω
2πc dλ
(5.7)
where λ is the wavelength of the light pulse and c is the light velocity. Reversing and
integrating Eq. 5.7, the phase shift, φ , can now be expressed as a function of the group
delay, tg, by
φ (λ ) = −2πc ∫
t g (λ )
λ2
70
dλ .
(5.8)
Chapter 5
Eq. 5.8 can be the basis for the phase ripple analysis of a CFBG in interference signal
broadening. Interference signal broadening is a common phenomenon observed and is
reported in Section 5.6 of this chapter.
In this chapter, an OCT system employing the all fibre optical delay line will
be constructed. Before that, initial characterization of the CFBGs’ group delay was
performed to acquire the dispersion properties and the group delay ripple (GDR) of
the gratings. Then the interference signal generated from CFBGs as an optical all fibre
delay line in an OCT system is reported and discussed.
5.3 Chirped Fibre Bragg Gratings Characterization
Using the in-house fibre Bragg grating fabrication technology, a pair of
linearly chirped FBGs was fabricated. The gratings are custom designed long CFBGs
[9] that are made from a hydrogen loaded and germanium doped core single mode
fibre at the operating wavelength of 1550 nm, of length 130 mm. The two gratings
were fabricated with a raised cosine apodisation profile over ~2% and 50%
respectively and thus yielded a top hat and a Gaussian-like shape function (Fig. 5.3)
The apodisation is crucial to prevent sidelobes and thus prevent the image degrading
in the OCT system. Both gratings were fabricated with a central wavelength at 1547
nm and the peak effective refractive index modulation was 3.7 x 10-4, while the chirp
rate in the gratings was kept at 3.846 nm/cm to yield the dispersion value of 25 ps/nm
1000
600
400
200
-20
0
-200
-30
Reflectivity (dB)
600
-10
400
-10
200
0
-20
Group delay (ps)
0
800
Group delay (ps)
Reflectivity (dB)
0
-200
-30
-400
-400
1520 1530 1540 1550 1560 1570
1520 1530 1540 1550 1560 1570
Wavelength (nm)
Wavelength (nm)
Fig. 5.3: Reflectivity spectrum and group delay for the top hat and Gaussian-shaped CFBGs. The top
hat grating is characterized from the short wavelength side and has a 50 nm FWHM bandwidth with 25
ps/nm dispersion. The Gaussian grating is characterized from the long wavelength side and has a 23 nm
FWHM bandwidth with -25 ps/nm dispersion.
71
Chapter 5
with 70% reflectivity. The characteristics of these CFBGs when used as a broadband
reflector and their group delay, are shown in Fig. 5.3. The reflected optical power of
the gratings is measured by the optical spectrum analyser (OSA). The group delay
property of these gratings is characterised using the modulation phase shift method
[10]. The setup for this characterization method is shown in Fig 5.4.
Grating under test
Tunable
laser
Intensity
modulator
EDFA
3 port
circulator
RF
Output
Photodiode
Network Analyser
(phase comparison)
RF
Input
Fig 5.4: Measurement setup for the modulation phase method.
The input light from a tunable laser is modulated by an optical intensity modulator
and the driving signal, with a frequency of f, is provided by the network analyser. The
modulated light is then launched into the grating under test. The reflected light signal
is then detected by a photodiode which later is introduced back into the network
analyser. The network analyser compared the phase difference between the received
signal and the driving electrical signal. As the wavelength of the tunable laser is
tuned, the phase difference among different wavelengths d φ (λ) is scanned and
analysed. The relative time delay at different wavelengths, which can be modified
from Eq. 5.7, is given by
t g (λ ) = −
=
λ2 dφ (λ )
⋅
2πc dλ
1 dφ (λ )
.
⋅
f 2π
72
(5.9)
Chapter 5
5.4 All Fibre Optical Delay Line Construction
Using these 2 gratings, a new all fibre optical delay line, as illustrated in Fig.
5.5, is designed. The top hat grating is UV glued at a point 1 cm away from both ends
of the grating, onto a piezo controlled stage at one end and onto a fixed stage at the
other end, thus giving a total of 150 mm length for straining (Fig 5.6). Since the top
hat and Gaussian gratings are arranged in a reverse cascaded orientation, the
dispersion value of the two gratings will then be compensated.
Top hat CFBG
1x2
3dB coupler
SOA
λshort
Isolator
2x2
3dB coupler
4 port
circulator
λlong
Gaussian CFBG
λshort
λlong
Reference arm
PC
PD
Sample
arm
4%
Fresnel
reflection
Oscilloscope
Fig. 5.5: Schematic diagram of the experimental set-up. SOA: semiconductor optical amplifier; PC:
polarisation controller; PD: photodetector; CFBG: chirped fibre Bragg grating. The constructed all
fibre optical delay line is highlighted in the shaded box.
Similarly with the original OCT setup, as reported in Section 3.2, the piezo
stage is driven with a triangular waveform to simulate a linear extension of the
grating, mimicking the real scanning mechanisms in the OCT system. A polarisation
controller was affixed to the sample arm to prevent the mismatch of polarisation that
arises from the induced birefringence in the constructed all fibre delay line. An
isolator was placed between the 1x2 3dB coupler and the 4 port circulator to prevent
the split signal from the light source from propagating in the reverse direction in the
reference arm. Instead of constructing focusing optics to scan a sample, the Fresnel
back-reflected signal from an end-cleaved fibre in the sample arm is used to match the
coherence length in both the sample and reference arms. Using this method, the
presence of bulk optics in the sample arm can be eliminated [11] and this proves that
73
Chapter 5
dispersion from the scanning optics in the sample arm does not contribute to the
defects of the interference signal.
1 cm
Light
spectrum
Fixed
stage
13 cm
λshort
1 cm
λlong
Piezo
stage
Index matching
fluid
Fig. 5.6: Schematic diagram of how the top hat CFBG is glued onto the fixed stage and piezo stage.
5.5 Interference Signal Generation
An SOA with a 1535.13 nm centre wavelength, + 0.41 dBm power and a
FWHM spectral bandwidth of 59.09 nm was used as the broadband, ASE light source.
Light travels in this reference arm in the following manner. Light will be split in half
at the 1x2 3 dB coupler. The one half of the split light will enter and circulate the 4
port circulator. The other half will be stopped at the isolator. While circulating
through the circulator ports, the light will be backreflected twice from the two
gratings attached to it. The first backreflection comes from the top hat grating that is
under a controlled stretch. The second grating will shape the light into a Gaussian-like
shape. Then the light will return to the 2x2 3 dB coupler, to interfere with light from
the sample arm, before hitting the photodetector.
Stretching the top hat grating will introduce a Bragg wavelength shift. Because
there is a dispersion property in the grating, the Bragg wavelength shift will translate
to a shift of time delay in the top hat grating. The Gaussian grating is a nearly
identical grating to the top hat grating except for its spectral shape. Thus, every point
in this grating is a mere reflection point to each wavelength component of the earlier
backreflected light from the top hat grating.
The reflection spectral profile resulting from the shaping of both gratings will
determine the shape of the interference signal as well. In this case, the final spectrum
is a Gaussian-like spectral shape incident on the photodetector, with the advantage of
giving suppression of unwanted sidelobes at its intereference signal shape. In order to
generate an interference signal, light from both the sample and the reference arm need
to interfere at the 2x2 3 dB coupler before being incident on the photodetector.
74
Chapter 5
5.6 Result and Discussion
The shift of Bragg wavelength, ΔλB, with strain can be expressed using
ΔλB=λB(1-pe) dl/Lgr (Eq. 5.4), where dl is the applied stretch on the strained grating
length, Lgr, and pe is the elasto-optic coefficient with a numerical value ~0.22. The top
hat grating is stretched by 162 μm; this will give rise to Bragg wavelength shift and
time delay of 1.33 nm and 33.37 ps respectively. This actually corresponds to a 3.45
mm depth for scanning which is sufficient for the penetration of light into the OCT
sample. The generated interference signal is shown in Fig. 5.7.
The coherence length, lc, which is the measure of the axial resolution in OCT,
is given by Eq. 2.14 in Section 2.1. The FWHM of the interference signal was
measured to be ~100 μm. Its coherence length is significantly broadened compared to
the theoretically predicted coherence length at 48 μm. The shape of the signal was an
approximate form of the interference signal in Fig. 4.6 (e).
The generated interference signal from the constructed all fibre optical delay
line, is anticipated to be a Gaussian-like shape, has been broadened by more than
twice the FWHM magnitude. It is speculated that the difference between the
experimental and the actual value can be attributed to the small dispersion imbalance
between the two CFBGs, in this new reference arm, together with the presence of
group delay ripple and phase ripple in the gratings. Light of broadband nature has
many wavelengths travelling at a group velocity. At large optical bandwidths, the
effect of dispersion on the interference signal, and hence the axial resolution, becomes
significant. Such a case has been reported where the light which enters the tissue, for
example, will undergo dispersion [12]. The different wavelengths return at slightly
different times. Hence the wider the bandwidth the higher the dispersion. In an
extreme case, if the dispersion is too great, the reference and sample arms will no
longer interfere and need to be compensated. After passing through the two gratings,
the effect of the gratings’ slight dispersion mismatch can cause the interference signal
to be widened and its frequency dispersion is then no longer symmetric. The selection
of an isolator, a 1x2 coupler and a circulator which are not entirely wavelength
flattened may have caused the Gaussian backreflected signal from the Gaussian
grating to be slightly distorted. This may have caused the interference signal to be
distorted and out of its original Gaussian shape.
75
Chapter 5
1.0
FWHM ~ 100 µm
Intensity (A.U.)
0.5
0.0
-0.5
-1.0
-200
-100
0
100
200
Delay (µm)
Fig. 5.7: Interference signal (solid) and the Gaussian envelope fit (dash) of a Fresnel backreflected
signal from the end-cleaved fibre in the constructed all fibre optical delay line OCT system.
But mainly, the presence of group delay ripple was thought to be the reason
for this broadening. In the future work, detailed in Chapter 6, a theoretical analysis of
the phase and group delay ripple in CFBGs will be suggested to find the cause. More
experimental investigation can be carried out in terms of numerical work to
understand the theoretical treatment of the phase ripple on the interference signal.
Future work can be concentrated on the simulation of the effect of these ripples on the
interference signal broadening.
76
Chapter 5
5.7 Doppler Frequency Observation
Going back to the same experimental setup as in Fig. 3.7, the Doppler
frequency of the new constructed delay line was measured. Three Doppler frequency
measurements were taken at 1 Hz, 2 Hz and 3 Hz piezo stage driving frequencies. The
recorded power incident on the photodetector (Model 1811, New Focus) was -21.36
dBm. This was lower than the earlier Doppler frequency observation in Section 3.4,
because of the multiple insertion loss introduced by circulators, gratings, couplers and
isolator in the delay line. The first Doppler frequency (Fig. 5.8 (a)) has the peak
centred at ~500 Hz. The Doppler frequencies in Fig 5.8 (b) have their peaks centred
~1 kHz. Lastly the Doppler frequencies in Fig 5.8 (c) have their peaks concentrated
in-between the peaks of ~1.5 kHz. The recorded parameters, on the Marconi RF
spectrum analyser, are as follows. All the three measurements are subject to only one
time sample averaging, as multiple times averaging will cause the peaks to gradually
decrease and diminish towards the end. The video bandwidth and resolution
bandwidth for the spectrum analyser was maintained at 87 Hz and 100 Hz
respectively. The signal attenuation into the spectrum analyser is 10 dB.
Again, similarly with the experimental setup carried out in Section 3.4,
frequency broadening is observed together with the increment of piezo driving
frequency. The reasons for the frequency broadening are also thought to be the same.
The interference signal generated has other frequency components besides the
Doppler frequency itself. Also, the unstable centre frequency, generated from the
function generator that is used to drive the piezo stage, may have contributed to the
frequency broadening too. This time, the distance of these peaks are no more 50 Hz
apart and thus it can be assured that the AC power line does not contribute to the
appearance of these multiple frequency peaks. By eliminating the mirror delay, the
newly constructed all fibre optical delay line would also not contribute any resonance
frequency to these multiple frequency peaks.
77
Chapter 5
Electrical Power (dBm)
-20
(a)
-40
-60
500 Hz
-80
-100
0
2000
4000
6000
8000 10000
Frequency (Hz)
Electrical Power (dBm)
-20
(b)
-40
1 kHz
-60
-80
-100
0
2000
4000
6000
8000 10000
Frequency (Hz)
Electrical Power (dBm)
-20
(c)
-40
Col 7 vs Col 8
-60
1.5 kHz
-80
-100
0
2000
4000
6000
8000 10000
Frequency (Hz)
Fig. 5.8: Doppler frequency of the all fibre optical delay line driven at (a) 1 Hz (b) 2 Hz and (c) 3 Hz.
The Doppler frequency broadening is observed together with the increment of the driving frequency.
78
Chapter 5
5.8 Summary
The construction of an all fibre optical delay line utilising a pair of CFBG has
been covered in this chapter. The gratings are fabricated with a top hat and Gaussian
shape-like function and each yields a dispersion value of 25 ps/nm. While cascading
the gratings in a reverse orientation, the dispersion values are canceled out. When the
top hat grating is stretched, an interference signal is obtained. The Bragg wavelength
shift of 1.33 nm, from a stretch of 162 µm to the top hat grating gave a penetration
depth scan of ~3.45 mm. This is more than adequate for depth scanning in OCT
applications. The fringe’s FWHM at 100 µm showed a broadening of twice the
theoretical value of 48 µm. The reason for this is believe to be due to the slight
dispersion mismatch in the OCT arms and the contribution of phase ripple in the
gratings. The measured Doppler frequencies are the same as with the earlier measured
frequency when using a flat mirror in the reference arm. This study has paved the way
for a more in-depth study of grating phase ripple in the future, as will be detailed in
Chapter 6.
79
Chapter 5
5.9 References
1.
L. P. Sanz, "High Speed Optical Delay Line for Optical Coherence
Tomography," in MSc Thesis(Technical University of Denmark, Roskilde, 2004).
2.
B. E. Bouma, and G. J. Tearney, Handbook of Optical Coherence
Tomography, (Marcel Dekker, New York, 2002).
3.
R. Kashyap, Fiber Bragg Gratings (Academic Press, New York, 1999).
4.
T. Erdogan, "Fiber grating spectra," Journal of Lightwave Technology 15,
1277-1294 (1997).
5.
B. H. Lee, T. J. Eom, E. Choi, G. Mudhana, and C. Lee, "Novel Optical Delay
Line for Optical Coherence Tomography System," Optical Review 10, 572-575
(2003).
6.
C. Yang, S. Yazdanfar, and J. Izatt, "Amplification of optical delay by use of
matched linearly chirped fiber Bragg gratings," Optics Letters 29, 685-687 (2004).
7.
A. D. Kersey, M. A. Davis, H. J. Patrick, M. LeBlanc, K. P. Koo, C. G.
Askins, M. A. Putnam, and E. J. Friebele, "Fiber grating sensors," Journal of
Lightwave Technology 15, 1442-1463 (1997).
8.
E. Choi, J. Na, S. Y. Ryu, and B. H. Lee, "Strained chirped fiber Bragg
gratings based all-fiber variable optical delay line for optical coherence tomography,"
Optical and Quantum Electronics 37, 1263-1276 (2005).
9.
M. Ibsen, M. K. Durkin, M. N. Zervas, A. B. Grudinin, and R. I. Laming,
"Custom Design of Long Chirped Bragg Gratings: Application to Gain-Flattening
Filter with Incorporated Dispersion Compensation," IEEE Photonics Technology
Letters 12, 498-500 (2000).
10.
Z. Zhang, "Passive and Active Bragg Gratings for Optical Networks," in Ph.D
Thesis, Optoelectronics Research Centre(University of Southampton, Southampton,
2007).
11.
E. Choi, J. Na, S. Y. Ryu, G. Mudhana, and B. H. Lee, "All-fiber variable
optical delay line for applications in optical coherence tomography: Feasibility study
for a novel delay line," Optics Express 13, 1334-1345 (2005).
12.
M. E. Brezinski, Chapter 5, Optical Coherence Tomography: Principles and
Applications (Elsevier, Burlington, 2006).
80
Chapter 6
Conclusions and Future Work
6.1 Summary
In this thesis, the development and construction of a time-domain OCT system
has been documented. A time-domain OCT system, utilising a mirror delay line at its
reference arm, has been designed. Its system performance has been demonstrated by
scanning a glass slide, a plastic sheet and a plastic laminator sample. The measured
fringe separation distance is in good agreement with the actual measured samples’
thickness. Thus, the unit can be used to perform a meaningful scan on any transparent
and thin samples. A simple and crude early stage OCT tomogram has been obtained
by scanning a mirror sample. A scan has been performed on an onion skin and small
features of the cellular structure, such as its borders and the middle of hollow cells,
can be determined. Overall, the constructed time-domain OCT system is able to
perform a crude meaningful scan on a simple biological sample.
This thesis also documented the work of a noise study of a broadband source
for application in the OCT system. The broadband source of interest in this case is the
high power, light-weight SC450 supercontinuum source (Fianium, UK). Interference
signals generated from this source were compared with other thermal-like sources;
i.e.; SOA and EDFA. A flat top and a Gaussian-like grating filter were used to shape
the broadband sources’ signal. The sources exhibit the exact Fourier transform
interference signal of their own filtered shape. The RIN value for the different filters,
using different broadband sources, has been compared and discussed too. For the
Gaussian shaped signal, the RIN value is slightly higher compared to the flat top
shaped signal. With the help of Dr. Anoma McCoy, the RIN value for the filtered
broadband sources have been calculated and predicted. Except for the SC450 source,
which is not a pure thermal source, the measured RIN value of each filtered SOA and
Chapter 6
EDFA source are in good agreement with their predicted values. The SC450 source
exhibits a high signal fluctuation and is noisier when compared to the SOA and
EDFA. This source is noisier due to two reasons; the factor of 2 increments in
intensity noise from its polarised signal and the nonlinear amplification of the
intensity noise in the pump laser due to complex nonlinear phenomena. Nevertheless,
the SC450 can still be utilised as a potential broadband source for OCT applications
based on the stability of its interference signal.
Later, realising that the constructed time-domain OCT system is very slow and
has very limited depth scanning penetration, the all fibre optical delay line is
constructed and then its operation is demonstrated. Using a pair of nearly identical
CFBGs, differing only in their coherence function shape, the gratings are cascaded in
a reverse orientation manner. This cancels out each other’s dispersion value. A nearly
Gaussian-like interference signal was produced when one of the CFBG was stretched.
A small stretch of the top hat grating amplified the depth scan penetration available at
the sample arm. The depth penetration is improved from the actual physical
displacement of 162 µm to 3.45 mm. Unfortunately the FWHM, or axial resolution, is
broadened by twice the magnitude of its theoretical value. The future work, as
highlighted in the next section, will discuss the methods and challenges to improve
this system.
6.2 Future Work
6.2.1 Noise Suppression
Limiting the noise is the key to achieving high resolution imaging in OCT.
Broadband CFBG filters centred at shorter wavelengths (1310 or 1100 nm), instead of
the 1550 nm wavelength, can be designed to characterize the interference signal and
its RIN. These wavelengths too are better suited in the OCT application due to the
lower scattering and absorption characteristics of the sample under test. Most of the
samples under OCT characterisation consist of inhomogeneous biological samples.
The nature of the supercontinuum source’s noise, either concerning the pump laser or
the PCF fundamental properties, is yet to be fully investigated. More experimental
work can be carried out to further establish the nature of the noise of the SC450
source before incorporating it into a high resolution OCT system. Once the parameters
82
Chapter 6
are known and the causes are determined, the possibility of reducing these noise
sources can be studied too in the future.
For the time being, by using the conventional balanced detection system, it
will be possible to increase the signal-to-noise ratio of the OCT system. The
commercial unit, model NirvanaTM, auto-balanced photodetector (New Focus, Inc.)
can provide the said solution, as depicted in Fig. 6.1. The suggested configuration, in
Fig. 6.1, will introduce the isolator between the C1 and C2 couplers to ensure that the
backreflected beam from the reference mirror will not increase the shot noise on the
photodetector. The equally split signal at the coupler can ensure that the intensity
noise is properly suppressed. In addition, the attenuator at one of the photodetector
arms can improve the detection efficiency. Attenuation at 3 dB is preferable, as the
common mode rejection ratio (CMRR) is maximised when power in the reference
input of the photodetector is twice that in the input signal.
Broadband
source
Sample
C1
Photodetector
Isolator
Mirror
+
C3
C2
Attenuator
Fig. 6.1: OCT system with balanced photodetector.
6.2.2 One micrometer Centre Wavelength Broadband Source
The considerations for choosing a light source need to take into account the
penetration depth and the absorption coefficient of water in a biological sample. It has
been shown that the dispersion zero of water is close to 1 µm wavelength [1] and the
effect of dispersion on the interference signal broadening can be minimised by
operating at this wavelength. Examples of broadband sources, that can emit at 1 μm
centre wavelength, are SLDs and some Ytterbium (Yb) doped fibre sources. An Yb
ASE source is a perfect example for the usage in the OCT system as it is readily
83
Chapter 6
1.0
Optical power (A.U.)
0.8
FWHM
~22.6 ± 0.1 nm
0.6
0.4
0.2
0.0
980
1000
1020
1040
1060
1080
Wavelength (nm)
Fig. 6.2: Emission spectrum of an in-house Yb ASE source.
available in the ORC grating lab. Its spectrum is shown in Fig. 6.2. Its centre
wavelength is at 1016.1 nm and the FWHM bandwidth is 22.6 nm. Its maximum
recorded power is 80 μW, or -11 dBm. At this optical bandwidth the axial resolution
for OCT will be 20.1 μm.
Take note that, when using the broad bandwidth spectrum centred at 1 μm, the
spectrum actually covers large portions of the detector sensitivity interface between
silicon and indium-germanium-arsenic (InGaAs) detectors. Germanium or Si–InGaAs
sandwiched detectors might be used, but these are less sensitive than silicon and
InGaAs detectors alone [2]. Fortunately, the InGaAs photodetector unit (D400FC,
Thorlabs) that was used throughout all the previous experiments, has a wide spectral
response, i.e., 800 nm – 1700 nm. This will be sufficient for a 1 μm central
wavelength signal, provided that the broadband source does not emit below 800 nm
wavelength.
Besides the photodetector, the output response of the OCT system is also
determined by the optical properties of the interferometer used. This covers the
wavelength-dependent losses and splitting ratios of the beamsplitter, as well as the
84
Chapter 6
cut-off wavelength of the single-mode fibres (SMFs) employed. A common error that
many researchers made was to mix together the different components with different
centre operating wavelengths in their setup. Each optical component is custom made
for its optimum operating wavelength. Special care needs to be taken especially when
splicing the fibres in the OCT setup. Due to different core and, hence, mode sizes,
when the light travels from 1.55 μm SMF into 1 μm SMF or vice versa, there will be a
significant splice loss. The minimum loss theoretically possible from this splice can
be shown using the following equation [3]
Minimum loss = 10 log10 (4r12 r22 (r12 + r22 ) −2 )
(6.1)
where r1 = 5.25 μm and r2 = 2.95 μm [4] are the mode field radii for the 1550 nm
SMF and 1060 nm SMF respectively. Therefore, for each fibre splice there will be a
minimum loss of ~ -1.37 dB. This problem can be avoided entirely by using 1 μm
operating wavelength components throughout the whole OCT setup.
6.2.3 Interference Signal Broadening Simulation
In Chapter 5, the interference signal is broaden from its theoretical axial
resolution of 48 μm to more than a factor of two at 100 μm. As speculated earlier, this
can be attributed to the phase ripple or noise properties in the gratings [5]. In Chapter
5 too, the group delay property (Fig. 5.3) is presented for both the gratings used in the
construction of the all fibre optical delay line. Fig. 6.3 shows the group delay ripples
(GDRs) in both gratings, which are obtained by taking a linear fit of the group delay
property and subtracting it from the measured group delay values. Random
fluctuations of up to ±10 ps and ±30 ps are observed for the top hat grating and
Gaussian grating respectively. At present, the tolerance value of GDR that will
influence the performance of the OCT system is still unknown. Nevertheless, the
presence of the GDR will become the essential study in the future to understand the
performance of an all fibre optical delay line OCT system.
The amplitude of the phase ripples is more indicative of the performance of
the chirped grating [6]. The phase ripple information in the grating can be obtained
using Eq. 5.8. This information can be extracted from the earlier group delay
85
20
60
10
30
GDR (ps)
GDR (ps)
Chapter 6
0
0
-10
-30
(a)
-20
1520 1530
(b)
-60
1520 1530
1540
1550
1560
1570
Wavelength (nm)
1540
1550
1560
1570
Wavelength (nm)
Fig. 6.3: Group delay ripple of (a) top hat grating and (b) Gaussian grating.
properties of the top hat and Gaussian gratings, from Fig. 5.3. The extracted phase
ripple properties are presented in Fig. 6.4.
Here, the study can be constrained to phase ripples, as opposed to group delay
or time ripple. By applying various periods and amplitudes onto the phase ripple, the
effect on the interference signal shaping can be interrogated. A relationship needs to
be established first between the phase noise and the stretching of the gratings in the
reference arm that can contribute towards the interference signal broadening. It is
hoped, in the future, that the nature of the phase ripple presence can be determined.
2000
0
0
-1000
Phase (rad)
Phase (rad)
-2000
-4000
-6000
-8000
-2000
-3000
-4000
-5000
-10000
-6000
(a)
-12000
1520
1530
1540
1550
1560
1570
Wavelength (nm)
(b)
-7000
1520 1530
1540
1550
1560
1570
Wavelength (nm)
Fig. 6.4: The phase ripple of (a) top hat grating and (b) Gaussian grating are simulated by employing
Eq. 5.8 using the group delay data in Fig. 5.3.
86
Chapter 6
6.2.4 Depth of Focus Improvement
The depth of focusing on the current OCT system can only focus at a meagre
depth of 18.59 μm. There are two ways to improve the depth of focus. Using a
technique called optical coherence microscopy (OCM), Izatt et al. [7] have produced
a series of two dimensional en face images using a high numerical aperture objective
lens. By adjusting the focus of the lens, the authors are able to penetrate further into
the sample and produce a series of depth scans. This is done by incorporating the light
rejection mechanism of a confocal microscopy setup and the coherence gating
technique of OCT. Another technique that can be used is to apply a form of dynamic
focusing of scanning optics at the sample arm [8]. This can be achieved by using a
special optical setup that can shift the focusing beam that is incident on the sample.
All this can be done without changing the optical path length in the interferometer’s
arms.
6.2.5 Fibre Optic Endoscope
The current OCT endoscope is bulky, expensive and full of moving
mechanical parts [9]. Fibre optic endoscopes are finding acceptance in biomedical
imaging because of their compactness and practicality in usage. Because of their
distinct miniscule size, this implies that they can be inserted into a blood vessel, for
example, for health monitoring purposes and many other suitable healthcare
applications. The idea of adapting this fibre optic endoscope into an OCT system has
been of great excitement and could revolutionize the medical industry. The endoscope
instrument itself can offer the inherent advantage of being safe and cheap, in addition
to its convenience of usage. Clinical applications of endoscopic OCT systems requires
probes that are quickly replaceable, cheap, compact and robust. A fibre optic
endoscope can be designed initially for proof of concept, before implementing it in
the OCT system. The current OCT setup is fibre based, thus enabling the possibility
of implementing fibre endoscope and catheter based imaging.
87
Chapter 6
6.3 References
1.
Y. Wang, J. S. Nelson, Z. Chen, B. J. Reiser, R. S. Chuck, and R. S. Windeler,
"Optimal wavelength for ultrahigh-resolution optical coherence tomography," Optics
Express 11, 1411-1417 (2003).
2.
W. Drexler, "Ultrahigh-resolution optical coherence tomography," Journal of
Biomedical Optics 9, 47-74 (2004).
3.
C. R. Pollack, Fundamentals of Optoelectronics (Irwin, Chicago, 1995).
4.
http://www.corning.com/opticalfiber.
5.
E. Choi, J. Na, S. Y. Ryu, and B. H. Lee, "Strained chirped fiber Bragg
gratings based all-fiber variable optical delay line for optical coherence tomography,"
Optical and Quantum Electronics 37, 1263-1276 (2005).
6.
J. F. Brennan, "Broadband fiber Bragg gratings for dispersion management,"
Journal of Optical and Fiber Communications Reports 2, 397-434 (2005).
7.
J. A. Izatt, M. D. Kulkarni, W. Hsing-Wen, K. Kobayashi, and M. V. Sivak,
Jr., "Optical coherence tomography and microscopy in gastrointestinal tissues," IEEE
Journal of Selected Topics in Quantum Electronics 2, 1017-1028 (1996).
8.
F. Lexer, C. K. Hitzenberger, W. Drexler, S. Molebny, H. Sattmann, M.
Sticker, and A. F. Fercher, "Dynamic coherent focus OCT with depth-independent
transversal resolution," Journal of Modern Optics 46, 541-553 (1999).
9.
G. J. Tearney, S. A. Boppart, B. E. Bouma, M. E. Brezinski, N. J. Weissman,
J. F. Southern, and J. G. Fujimoto, "Scanning single-mode fiber optic catheterendoscope for optical coherence tomography," Optics Letters 21, 543-545 (1996).
88
Appendix A
Start
Select MM4005
communication method
Enter triangular waveform
frequency and amplitude
Enter spot size (x) and
scanning distance (y)
N= y/x
MM4005 Motor ON
Waveform start
J+1
J=0
Reach waveform peak
MM4005 move x distance
Waveform ramp down
Reach waveform valley
Start recording interference signal
Waveform ramp up
Stack data in memory column J
J=N?
MM4005 Motor OFF
Waveform stop
Store memory on file
Request file name entry from user
END
89
Appendix B
90
91